An international mega-analysis of psychedelic drug effects on brain circuit function.

Code ↔ Paper

2 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 2 matches

[1] § Methods › Brain signal denoising ↔ ana8_uncorrected_data_pipelines_250820_NM.py, lines 138–215 · score 0.76 · noise components, independent components, ICA AROMA, aggressive, CSF, regresses
[2] § Methods › Brain signal denoising ↔ ana8_uncorrected_data_pipelines_250820_NM.py, lines 138–215 · score 0.71 · discrete cosine, component, masks, anatomical, WM, threshold

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Python · 4,756 lines · 195 KB · no license · 2 matches

import os
import numpy as np
import scipy as sc
import json
import glob
import pandas as pd
from sklearn.preprocessing import StandardScaler
from nilearn.input_data import NiftiLabelsMasker
from nilearn.signal import clean
from nilearn.interfaces.fmriprep import load_confounds, load_confounds_components
from nilearn.interfaces.fmriprep.load_confounds_components import _load_high_pass, _load_motion, _load_wm_csf, _load_global_signal, _load_compcor, _load_scrub
from nilearn import datasets as ds
import seaborn as sns
from matplotlib import pylab as plt
import pymc as pm
import arviz
from nilearn.interfaces.fmriprep import load_confounds_strategy
n_rois = 100 #200#, 400
for strategy in [
# ['motion', 'high_pass', 'compcor'],
# ['motion', 'high_pass', 'wm_csf', 'scrub'],
# ['wm_csf', 'high_pass', 'ica_aroma'],
# ['motion', 'high_pass', 'compcor', 'global_signal'],
# ['motion', 'high_pass', 'wm_csf', 'scrub', 'global_signal'],
['wm_csf', 'high_pass', 'ica_aroma', 'global_signal']
]:
# ['motion', 'high_pass', 'global_signal'],
# ['motion', 'high_pass', 'wm_csf', 'global_signal'],
# ['wm_csf', 'high_pass', 'ica_aroma', 'global_signal'],
# ['motion', 'high_pass', 'compcor', 'global_signal'],
# ['motion', 'high_pass', 'wm_csf', 'scrub', 'global_signal']]:
# ['motion', 'compcor'],
# ['motion', 'wm_csf', 'scrub'],
# ['aroma', 'wm_csf'],
# ['motion', 'compcor', 'global_signal'],
# ['motion', 'wm_csf', 'scrub', 'global_signal'],
# ['aroma', 'wm_csf', 'global_signal']]:
# ['motion', 'global_signal'], # for the roadout of fMRIPrep outputs
# ['wm_csf', 'global_signal'],
# ['ica_aroma', 'global_signal'],
# ['high_pass', 'global_signal'],
# ['scrub', 'global_signal'],
# ['compcor', 'global_signal'],
# ['motion'], # for the roadout of fMRIPrep outputs
# ['wm_csf'],
# ['ica_aroma'],
# ['high_pass'],
# ['scrub'],
# ['compcor']
CCs_all = []
FDs_all = []
conds_all = []
study_id_all = []
drug_all = [] # 0=psilocy (blue); 1=LSD (orange); 2=Aya (green); 3=mescaline (red), 4=DMT (purple)
drug_labels = np.array(['psilocybin', 'LSD', 'Aya', 'mescaline', 'DMT'])
plt.close('all')
strategy_suffix = '_'.join(strategy)
try:
os.mkdir(strategy_suffix)
except:
pass
# atlas = ds.fetch_atlas_schaefer_2018(n_rois=n_rois, yeo_networks=7)
atlas = ds.fetch_atlas_schaefer_2018(n_rois=n_rois, yeo_networks=17)
cort_labels = [l.decode() for l in atlas.labels]
masker = NiftiLabelsMasker(labels_img=atlas.maps)
masker.fit()
subcort_labels = list(pd.read_csv('_atlas_defs/Tian_subcortical38_labels.txt', header=None).values[:, 0])
claustrum_labels = list(pd.read_csv('_atlas_defs/Claustrum_labels.txt', header=None).values[:, 0])
cerebellum_labels = list(pd.read_csv('_atlas_defs/Cerebellum_labels.txt', header=None).values[:, 0])
# sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
full_labels = cort_labels + subcort_labels + claustrum_labels + cerebellum_labels
# construct subcort atlas
# sb_nii_paths = []
# for i in range(50):
# cur_nii = glob.glob('_scripts_data/collaboration_psychedelics_ROIs/subcortical/Tian_subcortex_*_%i.nii.gz' % (i + 1))
# if not len(cur_nii):
# pass
# else:
# sb_nii_paths.append(cur_nii)
# construct cerebellum atlas
from nilearn.image import math_img
import nibabel as nib
from nilearn.image import resample_img
cereb_nii_paths = []
cereb_nii_data = np.zeros( (97, 115, 97)).astype(int)
for i in range(17):
cur_nii = glob.glob('_scripts_data/collaboration_psychedelics_ROIs/cerebellum/rBuckner_cerebellum_%i.nii.gz' % (i + 1))
if not len(cur_nii):
pass
else:
print(cur_nii)
cereb_nii_paths.append(cur_nii[0])
cur_nii_bin = nib.load(cur_nii[0]).get_fdata().astype(int)
print(cur_nii_bin.sum())
cereb_nii_data += cur_nii_bin
head = nib.load(cereb_nii_paths[0]).header
aff = nib.load(cereb_nii_paths[0]).affine
cereb_nii = nib.Nifti1Image(cereb_nii_data, affine=aff, header=head)
cereb_nii.to_filename('atlas_cereb_Buckner17.nii.gz')
cereb_nii_r = resample_img(cereb_nii,
target_affine=masker.labels_img_.affine,
target_shape=masker.labels_img_.shape,
interpolation='nearest')
cereb_nii_r.to_filename('atlas_cereb_Buckner17_r.nii.gz')
masker_cereb = NiftiLabelsMasker(labels_img=cereb_nii_r)
masker_cereb.fit()
masker_cereb.labels_img_.to_filename('atlas_cereb_Buckner17_masker_cereb.nii.gz')
study_id = []
# net17_labels = []
# net7_label_names = np.array(['Vis', 'Som', 'DorsAtt', 'SalVent', 'Limbic',
# 'Cont', 'Default', 'TempPar'])
# for l in cort_labels:
# if '_Vis' in l:
# cur_l = 1
# elif '_Som' in l:
# cur_l = 2
# elif 'DorsAtt' in l:
# cur_l = 3
# elif '_SalVent' in l:
# cur_l = 4
# elif '_Limbic' in l:
# cur_l = 5
# elif '_Cont' in l:
# cur_l = 6
# elif 'Default' in l:
# cur_l = 7
# elif 'TempPar' in l:
# cur_l = 8
# else:
# print('error !')
# net17_labels.append(cur_l)
# net17_labels = np.array(net17_labels)
# net7_regcnt = np.bincount(net17_labels - 1)
net17_strs = [s.split('H_')[1].split('_')[0] for s in cort_labels]
net17_label_names = np.unique(net17_strs)
cort_dict = {net17_label_names[i]: (i+1) for i in np.arange(17)}
net17_labels = []
for cur_net_lab in net17_strs:
net17_labels.append(cort_dict[cur_net_lab])
net17_labels = np.array(net17_labels)
# Maastrict data
sub_strs = [p.split('/')[-1] for p in glob.glob('_data_uncorrected_fullcolumns/Maastricht_Psilocybin/sub-*')]
cond = pd.read_csv('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected/Maastricht/ROI_figures/conditions.txt', delimiter='\t')
def get_my_confounds_df(conf_t_path, conf_j_path, strategy):
conf_t = pd.read_csv(conf_t_path, delimiter='\t')
df_out = pd.DataFrame([])
new_cols = None
for cur_stra in strategy:
try:
if cur_stra == 'motion':
# full: translation/rotation + quadratic terms (24 parameters)
new_cols = getattr(load_confounds_components, '_load_motion')(conf_t, 'full')
elif cur_stra == 'wm_csf':
# basic: the averages in each mask (2 parameters)
new_cols = getattr(load_confounds_components, '_load_wm_csf')(conf_t, 'basic')
elif cur_stra == 'global_signal':
# basic: just the global signal (1 parameter)
new_cols = getattr(load_confounds_components, '_load_global_signal')(conf_t, 'basic')
elif cur_stra == 'ica_aroma':
# basic: aggressive/classic AROMA, only regress out the noise independent components
new_cols = getattr(load_confounds_components, '_load_ica_aroma')(conf_t, 'basic')
elif cur_stra == 'high_pass':
# add cosine columns to confound matrix to be regressed out
# adds discrete cosines transformation basis regressors to handle low-frequency signal drifts
new_cols = getattr(load_confounds_components, '_load_high_pass')(conf_t)
elif cur_stra == 'scrub':
# we "tag" rfMRI scans with 0/1 indicators that are then regressed out, instead of removing scans from time series
# fd_threhsold = 0.5 more regular threshold (Lior, various papers !!!)
new_cols = getattr(load_confounds_components, '_load_scrub')(conf_t, scrub=5, fd_threshold=0.5, std_dvars_threshold=3)
elif cur_stra == 'compcor':
# anat_combined: noise components calculated using a white matter and CSF combined anatomical mask
with open(conf_j_path, "rb") as f:
conf_j = json.load(f)
new_cols = getattr(load_confounds_components, '_load_compcor')(conf_t, conf_j, 'anat_combined', 5)
except:
print(f'!!! Exception when grapping {cur_stra} columns for {conf_j_path}')
if new_cols is not None:
df_out = pd.concat([df_out, new_cols], axis=1)
return df_out
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
print(sub_name)
conf_t_path = '/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Maastricht_Psilocybin/ROI_confounds/%s/func/%s_task-rest_desc-confounds_timeseries.tsv' % (sub_name, sub_name)
df = pd.read_csv(conf_t_path, delimiter='\t')
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
plt.title('FD (#volumes > 0.5mm out of %i total): Maastricht_Psilocybin' % len(df.framewise_displacement))
# plt.title('FD (#volumes > 0.5mm): Maastricht_Psilocybin\nimages > 0.2: %i/%i\nimages > 0.5: %i/%i\nimages > 0.8: %i/%i' % (
# sum(df.framewise_displacement >0.2), len(df.framewise_displacement),
# sum(df.framewise_displacement >0.5), len(df.framewise_displacement),
# sum(df.framewise_displacement >0.8), len(df.framewise_displacement)))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Maastricht_Psilocybin_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): Maastricht_Psilocybin' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Maastricht_Psilocybin_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): Maastricht_Psilocybin')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Maastricht_Psilocybin_dvars.png')
CCs = []
conds = []
n_skipped = 0
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Maastricht_Psilocybin/%s/func/%s_schaefer100_TS_noDenosing.csv' % (sub_name, sub_name), header=None)
subcort = pd.read_csv('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected/Maastricht/ROI_figures/%s/func/%s_subcortical38_TS_noDenosing.csv' % (sub_name, sub_name), header=None)
claustrum = pd.read_csv('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected/Maastricht/ROI_figures/%s/func/%s_claustrum_TS_noDenosing.csv' % (sub_name, sub_name), header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb = pd.read_csv('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected/Maastricht/ROI_figures/%s/func/%s_cerebellum_TS_noDenosing.csv' % (sub_name, sub_name), header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts))
conf_t_path = '/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Maastricht_Psilocybin/ROI_confounds/%s/func/%s_task-rest_desc-confounds_timeseries.tsv' % (sub_name, sub_name)
conf_j_path = '/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Maastricht_Psilocybin/ROI_confounds/%s/func/%s_task-rest_desc-confounds_timeseries.json' % (sub_name, sub_name)
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
conf = pd.DataFrame(np.nan_to_num(conf))
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
n_skipped += 1
print('Skipping subject: %s' % sub_name)
continue
# nii_fmri = masker.inverse_transform(sub_ts)
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
else:
continue
print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
cur_cond = cond[cond.Participant.values == int(sub_name[-2:])].Group_1isplacebo
if len(cur_cond) == 0:
continue
cur_cond = int(cur_cond.values == 2) # 1 == drug / 0 == placebo
study_id_all.append(0)
CCs.append(sub_CC)
conds.append(cur_cond)
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(cur_cond)
drug_all.append(0)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/maastrich_coverage.nii.gz')
CCs = np.array(CCs)
conds = np.array(conds)
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
CCs = np.nan_to_num(CCs)
out_mat = np.zeros((161, 161))
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_avg_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/maastricht_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/maastricht_cumabs_inter_regnorm_noabs.pdf')
# Palhano data: 18 subs, 2 conds; Aya dataset ! (mislabeled)
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Palhano_Psilocybin/*/sub-*/sub-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'preacute' if 'predosing' in sub_name else 'acute'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-ayahuasca%s_desc-confounds_timeseries.tsv' % cur_cond
sub_name = sub_strs[0].split('/')[-2]
print(sub_name)
df = pd.read_csv(conf_t_path, delimiter='\t')
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): palhano')
plt.title('FD (#volumes > 0.5mm out of %i total): palhano' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_palhano_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): palhano' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_palhano_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): palhano')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_palhano_dvars.png')
CCs = []
conds = []
n_skipped = 0
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_subcortical38_TS_noDenosing.csv'
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_claustrum_TS_noDenosing.csv'
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_cerebellum_TS_noDenosing.csv'
cereb = pd.read_csv(cereb_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
assert n_rois + 61 == sub_ts.shape[1]
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts))
cur_cond = 'preacute' if 'predosing' in sub_name else 'acute'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-ayahuasca%s_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path = conf_t_path.replace('.tsv', '.json')
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
n_skipped += 1
print('Skipping subject: %s' % sub_name)
continue
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
# signals=np.nan_to_num(sub_ts_z.values), standardize=False, np.nan_to_num(confounds=conf.iloc[:len(sub_ts_z)]))
else:
continue
# print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(1)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'acute'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'acute'))
drug_all.append(2)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/palhano_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Palhano_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Palhano_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Palhano_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/palhano_cumabs_inter_regnorm_noabs.pdf')
# Zurich data / LSD: 25 subs, 2 conds (Flora, Katrin)
del out_df
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Zurich_LSD_Psilocybin/output_LSD/sub-*/ses-*/func/sub-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'lsd' if 'ses-lsd' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
df = pd.read_csv(conf_t_path, delimiter='\t')
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): zurichlsd')
plt.title('FD (#volumes > 0.5mm out of %i total): zurichlsd' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_zurichlsd_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): zurichlsd' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Zurichlsd_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): zurichlsd')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_zurichlsd_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_subcortical38_TS_noDenosing.csv'
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_claustrum_TS_noDenosing.csv'
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_cerebellum_TS_noDenosing.csv'
cereb = pd.read_csv(cereb_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
assert n_rois + 61 == sub_ts.shape[1]
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts))
cur_cond = 'lsd' if 'ses-lsd' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
conf_j_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.json'
if not os.path.exists(conf_t_path):
print('Skipping this subject: confound file not found!')
continue
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
n_skipped += 1
print('Skipping subject: %s' % sub_name)
continue
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
else:
continue
print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(2)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'lsd'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'lsd'))
drug_all.append(1)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/zurichlsd_coverage.nii.gz')
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Zurich_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Zurich_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.5)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_regnorm.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_regnorm.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichLSD_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_regnorm_noabs.pdf')
# Zurich data / Psilo: 25 subs, 2 conds (Flora, Katrin)
del out_df
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Zurich_LSD_Psilocybin/output_Psilo/sub-*/ses-*/func/sub-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'lsd' if 'ses-lsd' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): zurichpsilo')
plt.title('FD (#volumes > 0.5mm out of %i total): zurichpsilo' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_zurichpsilo_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): zurichpsilo' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_zurichpsilo_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): zurichpsilo')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_zurichpsilo_dvars.png')
CCs = []
conds = []
n_skipped = 0
for sub_name in sub_strs:
# print(sub_name)
try:
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_subcortical38_TS_noDenosing.csv'
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_claustrum_TS_noDenosing.csv'
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_cerebellum_TS_noDenosing.csv'
cereb = pd.read_csv(cereb_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
assert n_rois + 61 == sub_ts.shape[1]
except:
print(f'CSV probably corrupted: {sub_name}')
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 'psi' if '-psi' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
conf_j_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.json'
if not os.path.exists(conf_t_path):
print('Skipping this subject: confound file not found!')
continue
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
n_skipped += 1
print('Skipping subject: %s' % sub_name)
continue
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
else:
continue
print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(3)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'psi'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'psi'))
drug_all.append(0)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/zurichpsilo_coverage.nii.gz')
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
if len(CCs) > 0:
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Zurich_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Zurich_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.5)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/ZurichPsilo_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/zurich_cumabs_inter_regnorm_noabs.pdf')
# continue # HACK !!! skipping Barrett for now
# Barrett data: 38 subs, 2 conds (Barrett, Manoj Doss)
del out_df
# import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Barrett_Psilocybin/*/*/*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'psilocybin' if 'Session5' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
sub_name = sub_strs[0].split('/')[-3]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): barrett')
plt.title('FD (#volumes > 0.5mm out of %i total): barrett' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_barrett_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): barrett' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_barrett_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): barrett')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_barrett_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_subcortical38_TS_noDenosing.csv'
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_claustrum_TS_noDenosing.csv'
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_cerebellum_TS_noDenosing.csv'
cereb = pd.read_csv(cereb_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
assert n_rois + 61 == sub_ts.shape[1]
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 'psilocybin' if 'Session5' in sub_name else 'plcb'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.tsv'
conf_j_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_desc-confounds_timeseries.json'
if not os.path.exists(conf_t_path):
print('Skipping this subject: confound file not found!')
continue
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
else:
continue
print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
# reg_avg_activity_bar.append(sub_ts.median(0).values)
# reg_avg_activity_bar_std.append(sub_ts.median(0).values)
#reg_avg_activity.append(sub_ts.mean(0).values)
study_id_all.append(4)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'psilocybin'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'psilocybin'))
drug_all.append(0)
# study_id.append(3)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/barrett_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_avg_CC_drug_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_df = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Barrett_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm_noabs.pdf')
# Roseman data: Psilocybin, 15 subjects
del out_df
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Roseman_Psilocybin/sub-*/ses-*/func/sub-*_ses-*_run*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'PSILO' if 'PSILO' in sub_name else 'plcb'
conf_t_path = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): Roseman_Psilocybin')
plt.title('FD (#volumes > 0.5mm out of %i total): Roseman_Psilocybin' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_Psilocybin_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): Roseman_Psilocybin' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_Psilocybin_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): Roseman_Psilocybin')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_Psilocybin_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_subcortical38_TS_noDenosing.csv'
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_claustrum_TS_noDenosing.csv'
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cereb_path = conf_path = sub_name.split('_schaefer100_TS')[0] + '_cerebellum_TS_noDenosing.csv'
cereb = pd.read_csv(cereb_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cereb], axis=1)
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 'PSILO' if 'PSILO' in sub_name else 'plcb'
conf_t_path = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path = conf_t_path.replace('.tsv', '.json')
if not os.path.exists(conf_t_path):
print('Skipping this subject: confound file not found!')
continue
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
else:
continue
print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
# reg_avg_activity_bar.append(sub_ts.median(0).values)
# reg_avg_activity_bar_std.append(sub_ts.median(0).values)
#reg_avg_activity.append(sub_ts.mean(0).values)
study_id_all.append(5)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'PSILO'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'PSILO'))
drug_all.append(0)
# study_id.append(3)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/rosemanpsilo_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_avg_CC_drug_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_drug_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb - CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb-drug.pdf')
from sklearn.linear_model import LinearRegression
out_df = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_df[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanPsilocybin_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm_noabs.pdf')
# Roseman data: LSD, 38 subjects
del out_df
# TODO: do the cross-corr in run1 and in run3 for a given subject first, then average the two
# to obtain a single cross-corr matrix for a given subject
# goal being to have one set of obsevation per subject (for lat statistical inference)
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Roseman_LSD/sub-*/ses-*/func/sub-*_ses-*_run-1_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'plcb' if '-plcb' in sub_name else 'LSD'
conf_t_path = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): Roseman_LSD')
plt.title('FD (#volumes > 0.5mm out of %i total): Roseman_LSD' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_LSD_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): Roseman_LSD' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_LSD_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): Roseman_LSD')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Roseman_LSD_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts1 = pd.read_csv(sub_name, header=None)
subcort1_path = sub_name.replace('schaefer%s' % n_rois, 'subcortical38')
subcort1 = pd.read_csv(subcort1_path, header=None)
claustrum1_path = sub_name.replace('schaefer%s' % n_rois, 'claustrum')
claustrum1 = pd.read_csv(claustrum1_path, header=None)
tmp = clean(
signals=claustrum1.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum1.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum1.iloc[:, [0, 3]] = tmp
cerebellum1_path = sub_name.replace('schaefer%s' % n_rois, 'cerebellum')
cerebellum1 = pd.read_csv(cerebellum1_path, header=None)
sub_ts1 = pd.concat([sub_ts1, subcort1, claustrum1, cerebellum1], axis=1)
assert sub_ts1.shape[1] == 161
sub_ts_z1 = pd.DataFrame(StandardScaler().fit_transform(sub_ts1), columns=full_labels)
cur_cond = 'plcb' if '-plcb' in sub_name else 'LSD'
conf_t_path1 = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path1 = conf_t_path.replace('.tsv', '.json')
if not os.path.exists(conf_t_path1):
print('Skipping this subject: confound file not found!')
continue
conf1 = get_my_confounds_df(conf_t_path1, conf_j_path1, strategy)
# motion correction
df = pd.read_csv(conf_t_path1, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if len(conf1) != 0:
conf1 = np.nan_to_num(conf1)
sub_ts_z_deconf1 = clean(
signals=sub_ts_z1.values, standardize=False, confounds=np.nan_to_num(conf1))
else:
continue
sub_CC1 = np.arctanh(np.corrcoef(sub_ts_z_deconf1.T))
sub_name2 = sub_name.replace('run-1', 'run-3')
sub_ts2 = pd.read_csv(sub_name2, header=None)
subcort2_path = subcort1_path.replace('run-1', 'run-3')
subcort2 = pd.read_csv(subcort2_path, header=None)
claustrum2_path = claustrum1_path.replace('run-1', 'run-3')
claustrum2 = pd.read_csv(claustrum2_path, header=None)
tmp = clean(
signals=claustrum2.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum2.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum2.iloc[:, [0, 3]] = tmp
cerebellum2_path = cerebellum1_path.replace('run-1', 'run-3')
cerebellum2 = pd.read_csv(cerebellum2_path, header=None)
sub_ts2 = pd.concat([sub_ts2, subcort2, claustrum2, cerebellum2], axis=1)
assert sub_ts2.shape[1] == 161
sub_ts_z2 = pd.DataFrame(StandardScaler().fit_transform(sub_ts2), columns=full_labels)
conf_t_path2 = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-3_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path2 = conf_t_path.replace('.tsv', '.json')
if not os.path.exists(conf_t_path2):
print('Skipping this subject: confound file not found!')
continue
conf2 = get_my_confounds_df(conf_t_path2, conf_j_path2, strategy)
# motion correction
df = pd.read_csv(conf_t_path2, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if len(conf2) != 0:
conf2 = np.nan_to_num(conf2)
sub_ts_z_deconf2 = clean(
signals=sub_ts_z2.values, standardize=False, confounds=np.nan_to_num(conf2))
else:
continue
sub_CC2 = np.arctanh(np.corrcoef(sub_ts_z_deconf2.T))
# reg_avg_activity_bar.append(sub_ts.median(0).values)
# reg_avg_activity_bar_std.append(sub_ts.median(0).values)
#reg_avg_activity.append(sub_ts.mean(0).values)
CCs.append(np.mean([sub_CC1, sub_CC2], axis=0)) # concatenating across run1 and run3 cross-corr matrices
conds.append(int(cur_cond == 'LSD'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(np.mean([sub_CC1, sub_CC2], axis=0)) # concatenating across run1 and run3 cross-corr matrices
conds_all.append(int(cur_cond == 'LSD'))
study_id_all.append(6)
drug_all.append(1)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/rosemanlsd_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_avg_CC_drug_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_drug_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb - CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb-drug.pdf')
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_avg_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/RosemanLSD_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm_noabs.pdf')
# Basel LSD data: 50 subs, 2 conds
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Basel_LSD_TS/sub*/*/func/sub-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'plcb' if 'ses-plcb' in sub_name else 'LSD'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
df = pd.read_csv(conf_t_path, delimiter='\t')
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): BaselLSD')
plt.title('FD (#volumes > 0.5mm out of %i total): Basel_LSD' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_BaselLSD_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): BaselLSD' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_BaselLSD_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): BaselLSD')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_BaselLSD_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = sub_name.replace('_schaefer%s' % n_rois, 'subcortical38')
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = sub_name.replace('_schaefer%s' % n_rois, 'claustrum')
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cerebellum_path = sub_name.replace('_schaefer%s' % n_rois, 'cerebellum')
cerebellum = pd.read_csv(cerebellum_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cerebellum], axis=1)
assert sub_ts.shape[1] == 161
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 'plcb' if 'ses-plcb' in sub_name else 'LSD'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
conf_j_path = conf_t_path.replace('.tsv', '.json')
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
# subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + 'subcortical38_TS_noDenosing.csv'
# subcort = pd.read_csv(subcort_path)
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
# signals=np.nan_to_num(sub_ts_z.values), standardize=False, np.nan_to_num(confounds=conf.iloc[:len(sub_ts_z)]))
else:
continue
# print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(7)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'LSD'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'LSD'))
drug_all.append(1)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/BaselLSD_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/BaselLSD_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/BaselLSD_cumabs_inter_regnorm_noabs.pdf')
# Basel LSD/Mesc/Psi data: 34 subs, 4 conds - LSD arm !!!
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Basel_LSD_Mescaline_Psilocybin/*/sub-*/*/func/sub-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'plcb'
if 'ses-LSD' in sub_name:
cur_cond = 'lsd'
elif 'ses-mesc' in sub_name:
cur_cond = 'mesc'
elif 'ses-psil' in sub_name:
cur_cond = 'psil'
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
print(sub_name.split('/')[-4])
df = pd.read_csv(conf_t_path, delimiter='\t')
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): Basel4cond')
plt.title('FD (#volumes > 0.5mm out of %i total): Basel4cond' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Basel4cond_FDabove0.5mm.png', dpi=300)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): Basel4cond' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Basel4cond_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): Basel4cond')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Basel4cond_dvars.png', dpi=300)
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = sub_name.replace('_schaefer%s' % n_rois, 'subcortical38')
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = sub_name.replace('_schaefer%s' % n_rois, 'claustrum')
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cerebellum_path = sub_name.replace('_schaefer%s' % n_rois, 'cerebellum')
cerebellum = pd.read_csv(cerebellum_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cerebellum], axis=1)
assert sub_ts.shape[1] == 161
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 0
if 'ses-LSD' in sub_name:
cur_cond = 1
elif 'ses-mesc' in sub_name:
continue # see below
elif 'ses-psil' in sub_name:
continue # see below
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
conf_j_path = conf_t_path.replace('.tsv', '.json')
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
# subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + 'subcortical38_TS_noDenosing.csv'
# subcort = pd.read_csv(subcort_path)
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
# signals=np.nan_to_num(sub_ts_z.values), standardize=False, np.nan_to_num(confounds=conf.iloc[:len(sub_ts_z)]))
else:
continue
# print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(8)
CCs.append(sub_CC)
conds.append(int(cur_cond))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond))
drug_all.append(1) # 0=psyloc (blue); 1=LSD (orange); 2=Aya (green)
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condLSD_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condLSD_cumabs_inter_regnorm_noabs.pdf')
# Basel LSD/Mesc/Psi data: 34 subs, 4 conds - Mesc arm !!!
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Basel_LSD_Mescaline_Psilocybin/*/sub-*/*/func/sub-*_schaefer100_TS_noDenosing.csv')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = sub_name.replace('_schaefer%s' % n_rois, 'subcortical38')
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = sub_name.replace('_schaefer%s' % n_rois, 'claustrum')
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cerebellum_path = sub_name.replace('_schaefer%s' % n_rois, 'cerebellum')
cerebellum = pd.read_csv(cerebellum_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cerebellum], axis=1)
assert sub_ts.shape[1] == 161
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 0
if 'ses-LSD' in sub_name:
continue
elif 'ses-mesc' in sub_name:
cur_cond = 1
elif 'ses-psil' in sub_name:
continue
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
conf_j_path = conf_t_path.replace('.tsv', '.json')
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
# subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + 'subcortical38_TS_noDenosing.csv'
# subcort = pd.read_csv(subcort_path)
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
# signals=np.nan_to_num(sub_ts_z.values), standardize=False, np.nan_to_num(confounds=conf.iloc[:len(sub_ts_z)]))
else:
continue
# print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# # stop
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(9)
CCs.append(sub_CC)
conds.append(int(cur_cond))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond))
drug_all.append(3) # 0=psyloc (blue); 1=LSD (orange); 2=Aya (green); 3=Mesc
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=cort_labels, yticklabels=cort_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condmesc_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condmesc_cumabs_inter_regnorm_noabs.pdf')
# Basel LSD/Mesc/Psi data: 34 subs, 4 conds - psilocybin arm !!!
import os
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Basel_LSD_Mescaline_Psilocybin/*/sub-*/*/func/sub-*_schaefer100_TS_noDenosing.csv')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
# print(sub_name)
sub_ts = pd.read_csv(sub_name, header=None)
subcort_path = sub_name.replace('_schaefer%s' % n_rois, 'subcortical38')
subcort = pd.read_csv(subcort_path, header=None)
claustrum_path = sub_name.replace('_schaefer%s' % n_rois, 'claustrum')
claustrum = pd.read_csv(claustrum_path, header=None)
tmp = clean(
signals=claustrum.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum.iloc[:, [0, 3]] = tmp
cerebellum_path = sub_name.replace('_schaefer%s' % n_rois, 'cerebellum')
cerebellum = pd.read_csv(cerebellum_path, header=None)
sub_ts = pd.concat([sub_ts, subcort, claustrum, cerebellum], axis=1)
assert sub_ts.shape[1] == 161
sub_ts_z = pd.DataFrame(StandardScaler().fit_transform(sub_ts), columns=full_labels)
cur_cond = 0
if 'ses-LSD' in sub_name:
continue
elif 'ses-mesc' in sub_name:
continue
elif 'ses-psil' in sub_name:
cur_cond = 1
conf_t_path = sub_name.split('_schaefer100_TS')[0] + '_task-rest_run-01_desc-confounds_timeseries.tsv'
conf_j_path = conf_t_path.replace('.tsv', '.json')
conf = get_my_confounds_df(conf_t_path, conf_j_path, strategy)
# motion correction
df = pd.read_csv(conf_t_path, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
# subcort_path = conf_path = sub_name.split('_schaefer100_TS')[0] + 'subcortical38_TS_noDenosing.csv'
# subcort = pd.read_csv(subcort_path)
if len(conf) != 0:
sub_ts_z_deconf = clean(
signals=np.nan_to_num(sub_ts_z.values), standardize=False, confounds=np.nan_to_num(conf))
# signals=np.nan_to_num(sub_ts_z.values), standardize=False, np.nan_to_num(confounds=conf.iloc[:len(sub_ts_z)]))
else:
continue
# print(f'{sub_name} has {len(conf.columns)} columns to noise clean!')
# sub_ts_z_deconf = pd.DataFrame(
# sub_ts_z_deconf, columns=cort_labels)
# # stop
# sub_CC = np.corrcoef(sub_ts_z_deconf.T)
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf.T))
study_id_all.append(10)
CCs.append(sub_CC)
conds.append(int(cur_cond))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond))
drug_all.append(0) # 0=psyloc (blue); 1=LSD (orange); 2=Aya (green); 3=Mesc
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_avg_CC_drug_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_avg_CC_drug_max.pdf')
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_intra.pdf')
# ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_inter.pdf')
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_inter_norm.pdf')
ax = pd.DataFrame(cum_intra_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net,
index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/Basel4condpsil_cumabs_inter_regnorm_%s.pdf' % strategy_suffix)
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Basel4condpsil_cumabs_inter_regnorm_noabs.pdf')
# Timmerman data: DMT, 20-ish subjects
del out_df
# TODO: do the cross-corr in run1 and in run3 for a given subject first, then average the two
# to obtain a single cross-corr matrix for a given subject
# goal being to have one set of obsevation per subject (for lat statistical inference)
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Timmermann_DMT/sub-*/ses-*/func/sub-*_ses-*schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
cur_cond = 'PCB' if 'PCB' in sub_name else 'DMT'
conf_t_path = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): timmermann_DMT')
plt.title('FD (#volumes > 0.5mm out of %i total): timmermann_DMT' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_timmermann_DMT_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): timmermann_DMT' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_timmermann_DMT_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): timmermann_DMT')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_timmermann_DMT_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
print(sub_name)
sub_ts1 = pd.read_csv(sub_name, header=None)
subcort1_path = sub_name.replace('_schaefer%s' % n_rois, 'subcortical38')
subcort1 = pd.read_csv(subcort1_path, header=None)
claustrum1_path = sub_name.replace('_schaefer%s' % n_rois, 'claustrum')
claustrum1 = pd.read_csv(claustrum1_path, header=None)
tmp = clean(
signals=claustrum1.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum1.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum1.iloc[:, [0, 3]] = tmp
cerebellum1_path = sub_name.replace('_schaefer%s' % n_rois, 'cerebellum')
cerebellum1 = pd.read_csv(cerebellum1_path, header=None)
sub_ts1 = pd.concat([sub_ts1, subcort1, claustrum1, cerebellum1], axis=1)
assert sub_ts1.shape[1] == 161
sub_ts_z1 = pd.DataFrame(StandardScaler().fit_transform(sub_ts1), columns=full_labels)
#print(sub_ts_z1)
cur_cond = 'PCB' if 'PCB' in sub_name else 'DMT'
conf_t_path1 = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-1_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path1 = conf_t_path.replace('.tsv', '.json')
# motion correction
df = pd.read_csv(conf_t_path1, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if not os.path.exists(conf_t_path1):
print('Skipping this subject: confound file not found!')
continue
conf1 = get_my_confounds_df(conf_t_path1, conf_j_path1, strategy)
if len(conf1) != 0:
conf1 = np.nan_to_num(conf1)
sub_ts_z_deconf1 = clean(
signals=sub_ts_z1.values, standardize=False, confounds=np.nan_to_num(conf1))
else:
continue
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf1.T))
# reg_avg_activity_bar.append(sub_ts.median(0).values)
# reg_avg_activity_bar_std.append(sub_ts.median(0).values)
#reg_avg_activity.append(sub_ts.mean(0).values)
study_id_all.append(11)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'DMT'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'DMT'))
drug_all.append(4) # 0=psyloc (blue); 1=LSD (orange); 2=Aya (green); 3=Mesc; 4=DMT
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_avg_CC_drug_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_drug_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb - CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb-drug.pdf')
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
#y = np.arctanh(CCs[:, i_roi, j_roi])
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_avg_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/timmermann_dmt_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm_noabs.pdf')
# Siegel data: 6 subjects
del out_df
# TODO: do the cross-corr in run1 and in run3 for a given subject first, then average the two
# to obtain a single cross-corr matrix for a given subject
# goal being to have one set of obsevation per subject (for lat statistical inference)
sub_strs = glob.glob('/Users/danilo/f/project_METAPSYCHO_BOLDpsychCONS/_data_uncorrected_fullcolumns/Siegel/sub-*_ses-*_schaefer100_TS_noDenosing.csv')
# QC: FD + dvars
study_FDs = []
study_dvars = []
for sub_name in sub_strs:
conf_t_path = sub_name.split('_ses-')[0] + '_ses-base_task-rest_run-01_desc-confounds_timeseries.tsv'
sub_name = sub_strs[0].split('/')[-4]
print(sub_name)
try:
df = pd.read_csv(conf_t_path, delimiter='\t')
except:
continue
cur_fd = (df.framewise_displacement > 0.5).sum()
study_FDs.append(cur_fd)
print(cur_fd)
cur_dvars = df.dvars.mean()
study_dvars.append(cur_dvars)
print(cur_dvars)
plt.figure()
plt.hist(study_FDs, bins=20)
# plt.title('FD (#volumes > 0.5mm): timmermann_DMT')
plt.title('FD (#volumes > 0.5mm out of %i total): Siegel' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Siegel_FDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(np.array(study_FDs) / len(df.framewise_displacement), bins=20)
plt.title('FD (fraction of volumes > 0.5mm out of %i total): Siegel' % len(df.framewise_displacement))
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Siegel_fracFDabove0.5mm.png', dpi=600)
plt.figure()
plt.hist(study_dvars, bins=20)
plt.title('DVARS (mean across TS): Siegel')
plt.savefig('_results_uncorrected_data_pipelines_ana8/QC_Siegel_dvars.png')
CCs = []
n_skipped = 0
conds = []
for sub_name in sub_strs:
print(sub_name)
sub_ts1 = pd.read_csv(sub_name, header=None)
subcort1_path = sub_name.replace('_schaefer%s' % n_rois, '-subcortical38')
subcort1 = pd.read_csv(subcort1_path, header=None)
claustrum1_path = sub_name.replace('_schaefer%s' % n_rois, '_claustrum')
claustrum1 = pd.read_csv(claustrum1_path, header=None)
tmp = clean(
signals=claustrum1.iloc[:, [0, 3]].to_numpy(),
confounds=claustrum1.iloc[:, [1, 2, 4, 5]].to_numpy(),
detrend=False, standardize=False, standardize_confounds=False, filter=False)
claustrum1.iloc[:, [0, 3]] = tmp
cerebellum1_path = sub_name.replace('_schaefer%s' % n_rois, '_cerebellum')
cerebellum1 = pd.read_csv(cerebellum1_path, header=None)
sub_ts1 = pd.concat([sub_ts1, subcort1, claustrum1, cerebellum1], axis=1)
assert sub_ts1.shape[1] == 161
sub_ts_z1 = pd.DataFrame(StandardScaler().fit_transform(sub_ts1), columns=full_labels)
#print(sub_ts_z1)
cur_cond = 'base' if 'base' in sub_name else 'psil'
conf_t_path1 = sub_name.split('_ses-')[0] + '_ses-%s_task-rest_run-01_desc-confounds_timeseries.tsv' % cur_cond
conf_j_path1 = conf_t_path.replace('.tsv', '.json')
# motion correction
df = pd.read_csv(conf_t_path1, delimiter='\t')
sub_motion_frac = (df.framewise_displacement > 0.5).sum() / len(df.framewise_displacement)
if sub_motion_frac > 0.15:
print('Skipping subject: %s' % sub_name)
n_skipped += 1
continue
if not os.path.exists(conf_t_path1):
print('Skipping this subject: confound file not found!')
continue
conf1 = get_my_confounds_df(conf_t_path1, conf_j_path1, strategy)
if len(conf1) != 0:
conf1 = np.nan_to_num(conf1)
sub_ts_z_deconf1 = clean(
signals=sub_ts_z1.values, standardize=False, confounds=np.nan_to_num(conf1))
else:
continue
sub_CC = np.arctanh(np.corrcoef(sub_ts_z_deconf1.T))
# reg_avg_activity_bar.append(sub_ts.median(0).values)
# reg_avg_activity_bar_std.append(sub_ts.median(0).values)
#reg_avg_activity.append(sub_ts.mean(0).values)
study_id_all.append(12)
CCs.append(sub_CC)
conds.append(int(cur_cond == 'psil'))
FDs_all.append(df.framewise_displacement.mean())
CCs_all.append(sub_CC)
conds_all.append(int(cur_cond == 'psil'))
drug_all.append(0) # 0=psyloc (blue); 1=LSD (orange); 2=Aya (green); 3=Mesc; 4=DMT
CCs = np.array(CCs)
conds = np.array(conds)
masker.inverse_transform(sub_ts_z1.mean(0)).to_filename('_results_uncorrected_data_pipelines_ana8/siegel_coverage.nii.gz')
CC_avg_plcb = CCs[conds == 0, :, :].mean(0)
CC_avg_drug = CCs[conds == 1, :, :].mean(0)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_avg_CC_plcb_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels, vmin=-1, vmax=+1)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_avg_CC_drug_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(16, 10))
sns.set(font_scale=0.4)
sns.heatmap(CC_avg_drug - CC_avg_plcb, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_avg_CC_drug-plcb_%s.pdf' % strategy_suffix)
# CC_avg_plcb_max = CCs[conds == 0, :, :].max(0)
# CC_avg_drug_max = CCs[conds == 1, :, :].max(0)
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug_max, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-1, vmax=+1)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_drug_max.pdf')
# import matplotlib.pylab as plt
# plt.figure(figsize=(16, 10))
# sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_plcb - CC_avg_drug, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/Barrett_avg_CC_plcb-drug.pdf')
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
CCs = np.nan_to_num(CCs)
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = CCs[:, i_roi, j_roi]
#y = np.arctanh(CCs[:, i_roi, j_roi])
X = conds[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_avg_lrcoef_%s.pdf' % strategy_suffix)
# import matplotlib.pylab as plt
# plt.figure(figsize=(20, 14))
# sns.set(font_scale=0.4)
# sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
# center=0, vmax=.35)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barret_lrcoef_viridis.pdf')
cum_intra_net = np.zeros((17))
cum_intra_net_cnt = np.zeros((17))
cum_inter_net = np.zeros((17))
cum_inter_net_cnt = np.zeros((17))
for i in range(n_rois): # rows
for j in range(n_rois): # columns
if i==j: # skip connections to self
continue
if net17_labels[i] == net17_labels[j]: # intra-network edge !
cum_intra_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_intra_net_cnt[net17_labels[i] - 1] += 1
if net17_labels[i] != net17_labels[j]: # inter-network edge !
cum_inter_net[net17_labels[i] - 1] += np.abs(out_df.iloc[i, j])
cum_inter_net_cnt[net17_labels[i] - 1] += 1
ax = pd.DataFrame(cum_intra_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Intra-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_avg_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
ax = pd.DataFrame(cum_inter_net, index=net17_label_names).plot.bar(fontsize=8, legend=False)
plt.ylabel('cumulative absolute beta coefficients', fontsize=10)
plt.title('Inter-network drug effects', fontsize=12)
plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter.pdf')
plt.savefig('_results_uncorrected_data_pipelines_ana8/siegel_cumabs_intra_regnorm_%s.pdf' % strategy_suffix)
# ax = pd.DataFrame(cum_intra_net / cum_intra_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_norm.pdf')
# ax = pd.DataFrame(cum_inter_net / cum_inter_net_cnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_norm.pdf')
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('cumulative absolute beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm.pdf')
# # no abs
# cum_intra_net = np.zeros((17))
# cum_intra_net_cnt = np.zeros((17))
# cum_inter_net = np.zeros((17))
# cum_inter_net_cnt = np.zeros((17))
# for i in range(n_rois): # rows
# for j in range(n_rois): # columns
# if i==j: # skip connections to self
# continue
# if net17_labels[i] == net17_labels[j]: # intra-network edge !
# cum_intra_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_intra_net_cnt[net17_labels[i] - 1] += 1
# if net17_labels[i] != net17_labels[j]: # inter-network edge !
# cum_inter_net[net17_labels[i] - 1] += out_df.iloc[i, j]
# cum_inter_net_cnt[net17_labels[i] - 1] += 1
# ax = pd.DataFrame(cum_intra_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Intra-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_intra_regnorm_noabs.pdf')
# ax = pd.DataFrame(cum_inter_net / net7_regcnt,
# index=net17_label_names).plot.bar(fontsize=8, legend=False)
# plt.ylabel('average beta coefficients (normalized)', fontsize=10)
# plt.title('Inter-network drug effects', fontsize=12)
# plt.tight_layout()
# plt.savefig('_results_uncorrected/barrett_cumabs_inter_regnorm_noabs.pdf')
plt.close('all')
##########################################################################################
# now running actual stats on all the aggregated data
##########################################################################################
FDs_all = np.array(FDs_all)
CCs_all = np.array(CCs_all)
conds_all = np.array(conds_all)
study_id_all = np.array(study_id_all)
drug_all = np.array(drug_all)
n_studies = len(np.unique(study_id_all))
from scipy.stats import ttest_rel, ttest_ind
# appeal / R3: do subjects move their heads differently during placebo vs. target ?
from scipy.stats import ttest_rel, ttest_ind
n = min(sum(conds_all==0),
sum(conds_all==1)
)
t, p = ttest_rel(
FDs_all[(conds_all==0)][:n],
FDs_all[(conds_all==1)][:n]
)
print('all drugs')
print(p)
all_avg_pvalue = 0
all_avg_tvalue = 0
a_all = []
b_all = []
for i_drug in np.unique(drug_all):
n = min(sum((conds_all==0) & (drug_all==i_drug)),
sum((conds_all==1) & (drug_all==i_drug))
)
a = FDs_all[(conds_all==0) & (drug_all==i_drug)][:n]
b = FDs_all[(conds_all==1) & (drug_all==i_drug)][:n]
t, p = ttest_rel(
a, # cond_all is drug status
b
)
print(drug_labels[i_drug])
print(p)
all_avg_pvalue += p
all_avg_tvalue += t
a_all += list(a)
b_all += list(b)
print('--- average p value ---')
print(p / (np.max(drug_all) + 1)) # p-value = 0.09263450080062638 + higher order; FDup2 p= ; FDup3 p=; FDup4, log p=; exp p=; 1/x p=, sqrt(p) =
print('--- average t value ---')
print(t / (np.max(drug_all) + 1)) # p-value = 0.09263450080062638 + higher order; FDup2 p= ; FDup3 p=; FDup4, log p=; exp p=; 1/x p=, sqrt(p) =
#boxplot
import seaborn as sns
X = pd.DataFrame((a_all, b_all)).T
X.columns = ['placebo', 'drug']
plt.figure(figsize=(6, 10))
sns.boxplot(X)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.title('framewise displacement', fontsize=18)
plt.savefig('_results_uncorrected_data_pipelines_ana8/FD_dotbydot.pdf')
# drug by drug with higher transformations
X = np.zeros((7, 5))
X_t = np.zeros((7, 5))
for i_drug in np.unique(drug_all):
n = min(sum((conds_all==0) & (drug_all==i_drug)),
sum((conds_all==1) & (drug_all==i_drug))
)
a = FDs_all[(conds_all==0) & (drug_all==i_drug)][:n]
b = FDs_all[(conds_all==1) & (drug_all==i_drug)][:n]
t, p = ttest_rel(
a, # cond_all is drug status
b
)
print(drug_labels[i_drug])
print(p)
X[0, i_drug] = p
X_t[0, i_drug] = t
t, p = ttest_rel(
a*a, # cond_all is drug status
b*b
)
print(drug_labels[i_drug])
print(p)
X[1, i_drug] = p
X_t[1, i_drug] = t
t, p = ttest_rel(
a*a*a, # cond_all is drug status
b*b*b
)
print(drug_labels[i_drug])
print(p)
X[2, i_drug] = p
X_t[2, i_drug] = t
t, p = ttest_rel(
a*a*a*a, # cond_all is drug status
b*b*b*b
)
print(drug_labels[i_drug])
print(p)
X[3, i_drug] = p
X_t[3, i_drug] = t
t, p = ttest_rel(
np.log(a), # cond_all is drug status
np.log(b)
)
print(drug_labels[i_drug])
print(p)
X[4, i_drug] = p
X_t[4, i_drug] = t
t, p = ttest_rel(
1 / np.array(a), # cond_all is drug status
1 / np.array(b)
)
print(drug_labels[i_drug])
print(p)
X[5, i_drug] = p
X_t[5, i_drug] = t
t, p = ttest_rel(
np.sqrt(a), # cond_all is drug status
np.sqrt(b)
)
print(drug_labels[i_drug])
print(p)
X[6, i_drug] = p
X_t[6, i_drug] = t
tranfs = ['x', 'x2', 'x3', 'x4' , 'log(x)', '1/x', 'sqrt(x)']
pX = pd.DataFrame(X, columns=drug_labels, index=tranfs)
pX_t = pd.DataFrame(X_t, columns=drug_labels, index=tranfs)
pX.to_csv('_results_uncorrected_data_pipelines_ana8/FD_drugbydrug_pvalues.csv')
pX_t = np.abs(pX_t)
pX_t.to_csv('_results_uncorrected_data_pipelines_ana8/FD_drugbydrug_tvalues.csv')
# pX.to_csv('_results_uncorrected_data_pipelines_ana8/FD_drugbydrug_pvalues.csv', float_format='%.4')
# pX_t.to_csv('_results_uncorrected_data_pipelines_ana8/FD_drugbydrug_tvalues.csv', float_format='%.4')
#acc = cross_val_score(estimator=LinearSVC(), X=FDs_all[:, None], y=conds_all[:, None], cv=KFold(n_splits=10, random_state=43, shuffle=True)); print(acc.mean()); print(acc.std()*2)
# 0.5984875983061102
# 0.15332339004087756
CCs_ravelized = CCs_all.reshape(CCs_all.shape[0], CCs_all.shape[1]*CCs_all.shape[1])
CCs_ravelized = np.nan_to_num(CCs_ravelized)
y = np.array(FDs_all > np.median(FDs_all))[:, None]
acc = cross_val_score(estimator=LinearSVC(), X=CCs_ravelized, y=y, cv=KFold(n_splits=10, random_state=43, shuffle=True)); print(acc.mean()); print(acc.std()*2)
# 0.501814882032668 -> so, no part int he brain, systematically distinguishes between high- versus low-motion participant connectomes in a robust way
# 0.15367835036406347 most holistic analyses that we can think of
for i_drug in np.arange(5):
acc = cross_val_score(estimator=LinearSVC(), X=CCs_ravelized[drug_all==i_drug], y=y[drug_all==i_drug], cv=KFold(n_splits=10, random_state=43, shuffle=True)); print(acc.mean()); print(acc.std()*2)
# none of 5 drugs, has above chance SVC classifier; when assessed drug by drug; that conclusion extends to any between-network and with-network effects as well
from sklearn.model_selection import cross_val_score, LeaveOneGroupOut, KFold
from sklearn.svm import LinearSVC
from scipy.stats import pearsonr
out_mat_placebo = np.zeros((161, 161))
out_mat_drug = np.zeros((161, 161))
for i_study in np.unique(study_id_all):
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
y = np.nan_to_num(CCs_all[:, i_roi, j_roi])
x = FDs_all
r, p = pearsonr(x[(conds_all==0) & (i_study==study_id_all)], y[(conds_all==0) & (i_study==study_id_all)])
out_mat_placebo[i_roi, j_roi] = r
r, p = pearsonr(x[(conds_all==1) & (i_study==study_id_all)], y[(conds_all==1) & (i_study==study_id_all)])
out_mat_drug[i_roi, j_roi] = r
out_df_placebo = pd.DataFrame(out_mat_placebo, columns=full_labels, index=full_labels)
out_df_placebo.to_csv('_results_uncorrected_data_pipelines_ana8/FD-FCmap_placebo_study%i_%s.csv' % (i_study+1, strategy_suffix))
out_df_drug = pd.DataFrame(out_mat_drug, columns=full_labels, index=full_labels)
out_df_drug.to_csv('_results_uncorrected_data_pipelines_ana8/FD-FCmap_drug_study%i_%s.csv' % (i_study+1, strategy_suffix))
print(ttest_ind(out_mat_placebo.ravel(), out_mat_drug.ravel()))
from scipy.stats import pearsonr
out_mat_placebo = np.zeros((161, 161))
out_mat_drug = np.zeros((161, 161))
for i_drug in np.unique(drug_all):
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
y = np.nan_to_num(CCs_all[:, i_roi, j_roi])
x = FDs_all
r, p = pearsonr(x[(conds_all==0) & (i_drug==drug_all)], y[(conds_all==0) & (i_drug==drug_all)])
out_mat_placebo[i_roi, j_roi] = r
r, p = pearsonr(x[(conds_all==1) & (i_drug==drug_all)], y[(conds_all==1) & (i_drug==drug_all)])
out_mat_drug[i_roi, j_roi] = r
out_df_placebo = pd.DataFrame(out_mat_placebo, columns=full_labels, index=full_labels)
out_df_placebo.to_csv('_results_uncorrected_data_pipelines_ana8/FD-FCmap_placebo_psyche%s_%s.csv' % (drug_labels[i_drug], strategy_suffix))
out_df_drug = pd.DataFrame(out_mat_drug, columns=full_labels, index=full_labels)
out_df_drug.to_csv('_results_uncorrected_data_pipelines_ana8/FD-FCmap_drug_psyche%s_%s.csv' % (drug_labels[i_drug], strategy_suffix))
print(ttest_ind(out_mat_placebo.ravel(), out_mat_drug.ravel()))
# revision: compute FD susceptibility map
from scipy.stats import pearsonr
out_mat = np.zeros((161, 161))
out_mat_p = np.zeros((161, 161))
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
y = np.nan_to_num(CCs_all[:, i_roi, j_roi])
x = FDs_all
r, p = pearsonr(x, y)
out_mat[i_roi, j_roi], out_mat_p[i_roi, j_roi] = r, p
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.close('all')
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5}, vmin=-1.0, vmax=1.0,
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_FD_suscept_%s.pdf' % strategy_suffix)
# out_mat[np.where(out_mat_p > (.05))] = 0
out_mat[np.where(out_mat_p > (.05/25921))] = 0
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5}, vmin=-1.0, vmax=1.0,
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_FD_suscept_p05_%s.pdf' % strategy_suffix)
for i_drug in np.unique(drug_all):
# revision: compute FD susceptibility map
from scipy.stats import pearsonr
out_mat = np.zeros((161, 161))
out_mat_p = np.zeros((161, 161))
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
y = np.nan_to_num(CCs_all[:, i_roi, j_roi])
x = FDs_all
r, p = pearsonr(x[drug_all==i_drug], y[drug_all==i_drug])
out_mat[i_roi, j_roi], out_mat_p[i_roi, j_roi] = r, p
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.close('all')
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5}, vmin=-1.0, vmax=1.0,
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_FD_suscept_%s_drug%s.pdf' % (strategy_suffix, drug_labels[i_drug]))
# out_mat[np.where(out_mat_p > (.05))] = 0
out_mat[np.where(out_mat_p > (.05/25921))] = 0
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5}, vmin=-1.0, vmax=1.0,
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_FD_suscept_p05_%s_drug%s.pdf' % (strategy_suffix, drug_labels[i_drug]))
# aggregate statistics: reg-reg conn
from sklearn.linear_model import LinearRegression
out_mat = np.zeros((161, 161))
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = np.nan_to_num(CCs_all[:, i_roi, j_roi])
X = conds_all[:, None]
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
import matplotlib.pylab as plt
plt.close('all')
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_lrcoef_%s.pdf' % strategy_suffix)
import matplotlib.pylab as plt
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(np.abs(out_df), cmap='viridis', square=True, cbar_kws={"shrink": 0.5},
center=0)#, vmax=.5)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_lrcoef_viridis_%s.pdf' % strategy_suffix)
# visu
from nilearn.input_data import NiftiLabelsMasker
masker = NiftiLabelsMasker(atlas.maps)
masker.fit()
regionwise_effects = np.abs(out_df).values.sum(1)[None, :]
out_nii1 = masker.inverse_transform(regionwise_effects)
out_nii2 = masker_cereb.inverse_transform(regionwise_effects[:, -17:])
out_nii = math_img('A + B', A=out_nii1, B=out_nii2)
out_nii.to_filename('_results_uncorrected_data_pipelines_ana8/13studies_avg_effects_%s.nii.gz' % strategy_suffix)
# aggregate statistics: reg-reg conn (PER DRUG)
from sklearn.linear_model import LinearRegression
for i_drug in np.unique(drug_all):
out_mat = np.zeros((161, 161))
for i_roi in range(161):
for j_roi in range(161):
if i_roi == j_roi:
continue
lr = LinearRegression(fit_intercept=True)
y = np.nan_to_num(CCs_all[drug_all == i_drug, i_roi, j_roi])
X = np.nan_to_num(conds_all[drug_all == i_drug, None])
lr.fit(X, y)
out_mat[i_roi, j_roi] = lr.coef_[0]
out_df = pd.DataFrame(out_mat, columns=full_labels, index=full_labels)
# visu
import matplotlib.pylab as plt
plt.close('all')
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
sns.heatmap(out_df, cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0)
plt.tight_layout()
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_lrcoef_%s_drug%s.pdf' % (strategy_suffix, drug_labels[i_drug]))
# from nilearn.input_data import NiftiLabelsMasker
# masker = NiftiLabelsMasker(atlas.maps)
# masker.fit()
# regionwise_effects = np.abs(out_df).values.sum(1)[None, :]
# out_nii = masker.inverse_transform(regionwise_effects)
from nilearn.input_data import NiftiLabelsMasker
masker = NiftiLabelsMasker(atlas.maps)
masker.fit()
regionwise_effects = np.abs(out_df).values.sum(1)[None, :]
out_nii1 = masker.inverse_transform(regionwise_effects)
out_nii2 = masker_cereb.inverse_transform(regionwise_effects[:, -17:])
out_nii = math_img('A + B', A=out_nii1, B=out_nii2)
out_nii.to_filename('_results_uncorrected_data_pipelines_ana8/13studies_avg_effects_%s_drug%s.nii.gz' % (strategy_suffix, drug_labels[i_drug]))
CC_avg_plcb = np.nanmean(CCs_all[(drug_all == i_drug) & (conds_all == 0), :, :], axis=0)
CC_avg_drug = np.nanmean(CCs_all[(drug_all == i_drug) & (conds_all == 1), :, :], axis=0)
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug - CC_avg_plcb , cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-0.25, vmax=+0.25)
sns.heatmap(np.nan_to_num(CC_avg_drug - CC_avg_plcb) , cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
pd.DataFrame(np.nan_to_num(CC_avg_drug - CC_avg_plcb), index=full_labels, columns=full_labels).to_csv('_results_uncorrected_data_pipelines_ana8/13studies_drug-plcb_%s_drug%s.csv' % (strategy_suffix, drug_labels[i_drug]))
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_drug-plcb_%s_drug%s.pdf' % (strategy_suffix, drug_labels[i_drug]))
#all
CC_avg_plcb = np.nanmean(CCs_all[(conds_all == 0), :, :], axis=0)
CC_avg_drug = np.nanmean(CCs_all[(conds_all == 1), :, :], axis=0)
plt.figure(figsize=(20, 14))
sns.set(font_scale=0.4)
# sns.heatmap(CC_avg_drug - CC_avg_plcb , cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
# center=0, xticklabels=cort_labels, yticklabels=cort_labels, vmin=-0.25, vmax=+0.25)
sns.heatmap(np.nan_to_num(CC_avg_drug - CC_avg_plcb) , cmap='RdBu_r', square=True, cbar_kws={"shrink": 0.5},
center=0, xticklabels=full_labels, yticklabels=full_labels)
plt.tight_layout()
pd.DataFrame(np.nan_to_num(CC_avg_drug - CC_avg_plcb), index=full_labels, columns=full_labels).to_csv('_results_uncorrected_data_pipelines_ana8/13studies_drug-plcb_%s_drugALL.csv' % (strategy_suffix))
plt.savefig('_results_uncorrected_data_pipelines_ana8/13studies_drug-plcb_%s_drugALL.pdf' % (strategy_suffix))
# # average CCs by network (to stablize the estimates that we input into the BHM), HPinterc
# # net-net
for cur_net1 in np.unique(net17_labels):
for cur_net2 in np.unique(net17_labels):
if cur_net1 == cur_net2:
continue
print('-' * 80)
cur_net_name1 = net17_label_names[cur_net1 - 1]
cur_net_name2 = net17_label_names[cur_net2 - 1]
print(cur_net_name1)
print('-- and --')
print(cur_net_name2)
net_inds1 = np.where(net17_labels == cur_net1)[0]
net_inds2 = np.where(net17_labels == cur_net2)[0]
y = CCs_all[:, net_inds1, :][:, :, net_inds2]
y = y.reshape((len(y), len(net_inds1)*len(net_inds2))).mean(1)
print(pd.DataFrame(y).describe())
# count 273.000000
# mean 0.340264
# std 0.134900
# min 0.078441
# 25% 0.238487
# 50% 0.314088
# 75% 0.428571
# max 0.750653
n_drugs = len(np.unique(drug_all))
pm_varnames = []
with pm.Model() as hierarchical_model:
hp_drugclass = pm.Normal('hp_drugclass', mu=0., sigma=1, shape=n_drugs)
hp_drugclass_priors = list(np.zeros(n_studies))
for j_drug in range(n_drugs):
for k_study in np.unique(study_id_all[drug_all == j_drug]):
hp_drugclass_priors[k_study] = hp_drugclass[j_drug]
alpha_drugclass = pm.Normal('alpha_drugclass', mu=hp_drugclass_priors, sigma=1, shape=n_studies)
# hp_cond = pm.Normal('hp_cond', mu=0., sigma=1, shape=1)
# alpha_cond = pm.Normal('alpha_cond', mu=hp_cond, sigma=1, shape=2)
# alpha_cond = pm.Normal('alpha_cond', mu=0, sigma=1, shape=2)
# beta_cond = pm.Normal('beta_cond', mu=0, sigma=1)
hp_beta = pm.Normal('hp_beta', mu=0., sigma=1, shape=1)
slope_betas = pm.Normal('slope_betas', mu=hp_beta, sigma=1, shape=n_drugs)
# beta_study_priors = list(np.zeros(n_studies))
# for j_drug in range(n_drugs):
# for k_study in np.unique(study_id_all[drug_all == j_drug]):
# beta_study_priors[k_study] = hp_beta[j_drug]
# beta_inter = pm.Normal('beta_inter_cond_drug', mu=beta_study_priors, sigma=2, shape=n_studies)
# reg_effect = alpha_cond[conds]
reg_effect = alpha_drugclass[study_id_all] + slope_betas[drug_all] * conds_all
eps = pm.HalfCauchy('eps', 5) # Model error
group_like = pm.Normal('beh_like', mu=reg_effect, sigma=eps, observed=y)
with hierarchical_model:
hierarchical_trace = pm.sample(draws=10000, n_init=5000, #init='advi',
chains=3, cores=9, progressbar=True,
random_seed=[1, 2, 3]) # one per chain needed
print(pm.summary(hierarchical_trace, hdi_prob=0.66))
OUT_DIR = '_results_uncorrected_data_pipelines_ana8'
output_name = 'net-net_HPinterc_effect_%s-%s' % (cur_net_name1, cur_net_name2)
t = hierarchical_trace
# for cur_roi in hierarchical_model.vars:
for cur_roi in hierarchical_model.named_vars:
from matplotlib.lines import Line2D
cur_roi = cur_roi
plt.close('all')
# THRESH = 0.5
# THRESH = 1.0
n_last_chains = 500
try:
suffix = ''
fig = arviz.plot_posterior(t, var_names=cur_roi, textsize=15.0,
hdi_prob=0.66)
# fig = pm.plot_posterior(t[-n_last_chains:], varnames=[cur_roi],
# credible_interval=0.95)
# try:
# for i_higher_cat in range(n_meta_cat):
# fig[i_higher_cat].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
plt.tight_layout()
plt.savefig('%s/%s_%s_posterior%s.png' % (OUT_DIR, output_name, cur_roi, suffix), dpi=150)
fig = arviz.plot_trace(t, var_names=[cur_roi])
# max_abs_mode = np.max(np.abs(hierarchical_trace2[-n_last_chains:][cur_roi].mean(0)))
# fig = pm.traceplot(t[-n_last_chains:], varnames=[cur_roi])
# max_abs_mode = np.max(np.abs(t[-n_last_chains:][cur_roi].mean(0)))
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# try:
# fig[1][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# try:
# if max_abs_mode < THRESH and not 'nuisance' in cur_roi:
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# post_lines = fig[0][0].get_lines()
# custom_lines = [Line2D([0], [0], color=l.get_c(), lw=4) for l in post_lines]
# if subgroup_labels is None:
# subgroup_labels = ['subgroup %i' % i for i in range(len(custom_lines))]
# fig[0][0].legend(custom_lines, subgroup_labels, loc='upper left', prop={'size': 7.5})
plt.tight_layout()
plt.savefig('%s/%s_%s%s.png' % (OUT_DIR, output_name, cur_roi,
suffix), dpi=150)
# plt.savefig('%s/%s_%s.png' % (OUT_DIR, output_name, cur_roi), dpi=150)
except:
pass
# # average CCs by network (to stablize the estimates that we input into the BHM), HPinterc
# # net-net (across drug average)
for cur_net1 in np.unique(net17_labels):
for cur_net2 in np.unique(net17_labels):
if cur_net1 == cur_net2:
continue
print('-' * 80)
cur_net_name1 = net17_label_names[cur_net1 - 1]
cur_net_name2 = net17_label_names[cur_net2 - 1]
print(cur_net_name1)
print('-- and --')
print(cur_net_name2)
net_inds1 = np.where(net17_labels == cur_net1)[0]
net_inds2 = np.where(net17_labels == cur_net2)[0]
y = CCs_all[:, net_inds1, :][:, :, net_inds2]
y = y.reshape((len(y), len(net_inds1)*len(net_inds2))).mean(1)
print(pd.DataFrame(y).describe())
# count 273.000000
# mean 0.340264
# std 0.134900
# min 0.078441
# 25% 0.238487
# 50% 0.314088
# 75% 0.428571
# max 0.750653
n_drugs = len(np.unique(drug_all))
pm_varnames = []
with pm.Model() as hierarchical_model:
hp_drugclass = pm.Normal('hp_drugclass', mu=0., sigma=1, shape=n_drugs)
hp_drugclass_priors = list(np.zeros(n_studies))
for j_drug in range(n_drugs):
for k_study in np.unique(study_id_all[drug_all == j_drug]):
hp_drugclass_priors[k_study] = hp_drugclass[j_drug]
alpha_drugclass = pm.Normal('alpha_drugclass', mu=hp_drugclass_priors, sigma=1, shape=n_studies)
# hp_cond = pm.Normal('hp_cond', mu=0., sigma=1, shape=1)
# alpha_cond = pm.Normal('alpha_cond', mu=hp_cond, sigma=1, shape=2)
# alpha_cond = pm.Normal('alpha_cond', mu=0, sigma=1, shape=2)
# beta_cond = pm.Normal('beta_cond', mu=0, sigma=1)
#hp_beta = pm.Normal('hp_beta', mu=0., sigma=1, shape=1)
slope_betas = pm.Normal('slope_betas', mu=0, sigma=1, shape=1)
# beta_study_priors = list(np.zeros(n_studies))
# for j_drug in range(n_drugs):
# for k_study in np.unique(study_id_all[drug_all == j_drug]):
# beta_study_priors[k_study] = hp_beta[j_drug]
# beta_inter = pm.Normal('beta_inter_cond_drug', mu=beta_study_priors, sigma=2, shape=n_studies)
# reg_effect = alpha_cond[conds]
reg_effect = alpha_drugclass[study_id_all] + slope_betas * conds_all
eps = pm.HalfCauchy('eps', 5) # Model error
group_like = pm.Normal('beh_like', mu=reg_effect, sigma=eps, observed=y)
with hierarchical_model:
hierarchical_trace = pm.sample(draws=10000, n_init=5000, #init='advi',
chains=3, cores=9, progressbar=True,
random_seed=[1, 2, 3]) # one per chain needed
print(pm.summary(hierarchical_trace, hdi_prob=0.66))
OUT_DIR = '_results_uncorrected_data_pipelines_ana8'
output_name = 'net-net_HPinterc_effect_%s-%s_across' % (cur_net_name1, cur_net_name2)
t = hierarchical_trace
for cur_roi in hierarchical_model.named_vars:
from matplotlib.lines import Line2D
cur_roi = cur_roi
plt.close('all')
# THRESH = 0.5
# THRESH = 1.0
n_last_chains = 500
try:
suffix = ''
fig = arviz.plot_posterior(t, var_names=cur_roi, textsize=15.0,
hdi_prob=0.66)
# fig = pm.plot_posterior(t[-n_last_chains:], varnames=[cur_roi],
# credible_interval=0.95)
# try:
# for i_higher_cat in range(n_meta_cat):
# fig[i_higher_cat].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
plt.tight_layout()
plt.savefig('%s/%s_%s_posterior%s.png' % (OUT_DIR, output_name, cur_roi, suffix), dpi=150)
fig = arviz.plot_trace(t, var_names=[cur_roi])
# max_abs_mode = np.max(np.abs(hierarchical_trace2[-n_last_chains:][cur_roi].mean(0)))
# fig = pm.traceplot(t[-n_last_chains:], varnames=[cur_roi])
# max_abs_mode = np.max(np.abs(t[-n_last_chains:][cur_roi].mean(0)))
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# try:
# fig[1][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# try:
# if max_abs_mode < THRESH and not 'nuisance' in cur_roi:
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# post_lines = fig[0][0].get_lines()
# custom_lines = [Line2D([0], [0], color=l.get_c(), lw=4) for l in post_lines]
# if subgroup_labels is None:
# subgroup_labels = ['subgroup %i' % i for i in range(len(custom_lines))]
# fig[0][0].legend(custom_lines, subgroup_labels, loc='upper left', prop={'size': 7.5})
plt.tight_layout()
plt.savefig('%s/%s_%s%s.png' % (OUT_DIR, output_name, cur_roi,
suffix), dpi=150)
# plt.savefig('%s/%s_%s.png' % (OUT_DIR, output_name, cur_roi), dpi=150)
except:
pass
# # average CCs by network (to stablize the estimates that we input into the BHM), HPinterc
# # net-THA-D
net_THAD_labels = np.array([s.startswith('THA-D') * 20 for s in full_labels])
for cur_net1 in np.unique(net17_labels):
for cur_net2 in np.unique(net_THAD_labels):
if cur_net2 == 0:
continue
print('-' * 80)
cur_net_name1 = net17_label_names[cur_net1 - 1]
cur_net_name2 = 'THA-D'
print(cur_net_name1)
print('-- and --')
print(cur_net_name2)
net_inds1 = np.where(net17_labels == cur_net1)[0]
net_inds2 = np.where(net_THAD_labels == cur_net2)[0]
y = CCs_all[:, net_inds1, :][:, :, net_inds2]
y = y.reshape((len(y), len(net_inds1)*len(net_inds2))).mean(1)
print(pd.DataFrame(y).describe())
# count 273.000000
# mean 0.340264
# std 0.134900
# min 0.078441
# 25% 0.238487
# 50% 0.314088
# 75% 0.428571
# max 0.750653
n_drugs = len(np.unique(drug_all))
pm_varnames = []
with pm.Model() as hierarchical_model:
hp_drugclass = pm.Normal('hp_drugclass', mu=0., sigma=1, shape=n_drugs)
hp_drugclass_priors = list(np.zeros(n_studies))
for j_drug in range(n_drugs):
for k_study in np.unique(study_id_all[drug_all == j_drug]):
hp_drugclass_priors[k_study] = hp_drugclass[j_drug]
alpha_drugclass = pm.Normal('alpha_drugclass', mu=hp_drugclass_priors, sigma=1, shape=n_studies)
# hp_cond = pm.Normal('hp_cond', mu=0., sigma=1, shape=1)
# alpha_cond = pm.Normal('alpha_cond', mu=hp_cond, sigma=1, shape=2)
# alpha_cond = pm.Normal('alpha_cond', mu=0, sigma=1, shape=2)
# beta_cond = pm.Normal('beta_cond', mu=0, sigma=1)
hp_beta = pm.Normal('hp_beta', mu=0., sigma=1, shape=1)
slope_betas = pm.Normal('slope_betas', mu=hp_beta, sigma=1, shape=n_drugs)
# beta_study_priors = list(np.zeros(n_studies))
# for j_drug in range(n_drugs):
# for k_study in np.unique(study_id_all[drug_all == j_drug]):
# beta_study_priors[k_study] = hp_beta[j_drug]
# beta_inter = pm.Normal('beta_inter_cond_drug', mu=beta_study_priors, sigma=2, shape=n_studies)
# reg_effect = alpha_cond[conds]
reg_effect = alpha_drugclass[study_id_all] + slope_betas[drug_all] * conds_all
eps = pm.HalfCauchy('eps', 5) # Model error
group_like = pm.Normal('beh_like', mu=reg_effect, sigma=eps, observed=y)
with hierarchical_model:
hierarchical_trace = pm.sample(draws=10000, n_init=5000, #init='advi',
chains=3, cores=9, progressbar=True,
random_seed=[1, 2, 3]) # one per chain needed
print(pm.summary(hierarchical_trace, hdi_prob=0.66))
OUT_DIR = '_results_uncorrected_data_pipelines_ana8'
output_name = 'net-THAD_HPinterc_effect_%s-%s' % (cur_net_name1, cur_net_name2)
t = hierarchical_trace
for cur_roi in hierarchical_model.named_vars:
from matplotlib.lines import Line2D
cur_roi = cur_roi
plt.close('all')
# THRESH = 0.5
# THRESH = 1.0
n_last_chains = 500
try:
suffix = ''
fig = arviz.plot_posterior(t, var_names=cur_roi, textsize=15.0,
hdi_prob=0.66)
# fig = pm.plot_posterior(t[-n_last_chains:], varnames=[cur_roi],
# credible_interval=0.95)
# try:
# for i_higher_cat in range(n_meta_cat):
# fig[i_higher_cat].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
plt.tight_layout()
plt.savefig('%s/%s_%s_posterior%s.png' % (OUT_DIR, output_name, cur_roi, suffix), dpi=150)
fig = arviz.plot_trace(t, var_names=[cur_roi])
# max_abs_mode = np.max(np.abs(hierarchical_trace2[-n_last_chains:][cur_roi].mean(0)))
# fig = pm.traceplot(t[-n_last_chains:], varnames=[cur_roi])
# max_abs_mode = np.max(np.abs(t[-n_last_chains:][cur_roi].mean(0)))
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# try:
# fig[1][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# try:
# if max_abs_mode < THRESH and not 'nuisance' in cur_roi:
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# post_lines = fig[0][0].get_lines()
# custom_lines = [Line2D([0], [0], color=l.get_c(), lw=4) for l in post_lines]
# if subgroup_labels is None:
# subgroup_labels = ['subgroup %i' % i for i in range(len(custom_lines))]
# fig[0][0].legend(custom_lines, subgroup_labels, loc='upper left', prop={'size': 7.5})
plt.tight_layout()
plt.savefig('%s/%s_%s%s.png' % (OUT_DIR, output_name, cur_roi,
suffix), dpi=150)
# plt.savefig('%s/%s_%s.png' % (OUT_DIR, output_name, cur_roi), dpi=150)
except:
pass
# # average CCs by network (to stablize the estimates that we input into the BHM), HPinterc
# # net-THA-D (across drug average)
net_THAD_labels = np.array([s.startswith('THA-D') * 20 for s in full_labels])
for cur_net1 in np.unique(net17_labels):
for cur_net2 in np.unique(net_THAD_labels):
if cur_net2 == 0:
continue
print('-' * 80)
cur_net_name1 = net17_label_names[cur_net1 - 1]
cur_net_name2 = 'THA-D'
print(cur_net_name1)
print('-- and --')
print(cur_net_name2)
net_inds1 = np.where(net17_labels == cur_net1)[0]
net_inds2 = np.where(net_THAD_labels == cur_net2)[0]
y = CCs_all[:, net_inds1, :][:, :, net_inds2]
y = y.reshape((len(y), len(net_inds1)*len(net_inds2))).mean(1)
print(pd.DataFrame(y).describe())
# count 273.000000
# mean 0.340264
# std 0.134900
# min 0.078441
# 25% 0.238487
# 50% 0.314088
# 75% 0.428571
# max 0.750653
n_drugs = len(np.unique(drug_all))
pm_varnames = []
with pm.Model() as hierarchical_model:
hp_drugclass = pm.Normal('hp_drugclass', mu=0., sigma=1, shape=n_drugs)
hp_drugclass_priors = list(np.zeros(n_studies))
for j_drug in range(n_drugs):
for k_study in np.unique(study_id_all[drug_all == j_drug]):
hp_drugclass_priors[k_study] = hp_drugclass[j_drug]
alpha_drugclass = pm.Normal('alpha_drugclass', mu=hp_drugclass_priors, sigma=1, shape=n_studies)
# hp_cond = pm.Normal('hp_cond', mu=0., sigma=1, shape=1)
# alpha_cond = pm.Normal('alpha_cond', mu=hp_cond, sigma=1, shape=2)
# alpha_cond = pm.Normal('alpha_cond', mu=0, sigma=1, shape=2)
# beta_cond = pm.Normal('beta_cond', mu=0, sigma=1)
#hp_beta = pm.Normal('hp_beta', mu=0., sigma=1, shape=1)
slope_betas = pm.Normal('slope_betas', mu=0, sigma=1, shape=1)
# beta_study_priors = list(np.zeros(n_studies))
# for j_drug in range(n_drugs):
# for k_study in np.unique(study_id_all[drug_all == j_drug]):
# beta_study_priors[k_study] = hp_beta[j_drug]
# beta_inter = pm.Normal('beta_inter_cond_drug', mu=beta_study_priors, sigma=2, shape=n_studies)
# reg_effect = alpha_cond[conds]
reg_effect = alpha_drugclass[study_id_all] + slope_betas * conds_all
eps = pm.HalfCauchy('eps', 5) # Model error
group_like = pm.Normal('beh_like', mu=reg_effect, sigma=eps, observed=y)
with hierarchical_model:
hierarchical_trace = pm.sample(draws=10000, n_init=5000, #init='advi',
chains=3, cores=9, progressbar=True,
random_seed=[1, 2, 3]) # one per chain needed
print(pm.summary(hierarchical_trace, hdi_prob=0.66))
OUT_DIR = '_results_uncorrected_data_pipelines_ana8'
output_name = 'net-THAD_HPinterc_effect_%s-%s_across' % (cur_net_name1, cur_net_name2)
t = hierarchical_trace
for cur_roi in hierarchical_model.named_vars:
from matplotlib.lines import Line2D
cur_roi = cur_roi
plt.close('all')
# THRESH = 0.5
# THRESH = 1.0
n_last_chains = 500
try:
suffix = ''
fig = arviz.plot_posterior(t, var_names=cur_roi, textsize=15.0,
hdi_prob=0.66)
# fig = pm.plot_posterior(t[-n_last_chains:], varnames=[cur_roi],
# credible_interval=0.95)
# try:
# for i_higher_cat in range(n_meta_cat):
# fig[i_higher_cat].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
plt.tight_layout()
plt.savefig('%s/%s_%s_posterior%s.png' % (OUT_DIR, output_name, cur_roi, suffix), dpi=150)
fig = arviz.plot_trace(t, var_names=[cur_roi])
# max_abs_mode = np.max(np.abs(hierarchical_trace2[-n_last_chains:][cur_roi].mean(0)))
# fig = pm.traceplot(t[-n_last_chains:], varnames=[cur_roi])
# max_abs_mode = np.max(np.abs(t[-n_last_chains:][cur_roi].mean(0)))
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# try:
# fig[1][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# try:
# if max_abs_mode < THRESH and not 'nuisance' in cur_roi:
# fig[0][0].set_xlim(-THRESH, THRESH) # make plots more comparable
# except:
# pass
# post_lines = fig[0][0].get_lines()
# custom_lines = [Line2D([0], [0], color=l.get_c(), lw=4) for l in post_lines]
# if subgroup_labels is None:
# subgroup_labels = ['subgroup %i' % i for i in range(len(custom_lines))]
# fig[0][0].legend(custom_lines, subgroup_labels, loc='upper left', prop={'size': 7.5})
plt.tight_layout()
plt.savefig('%s/%s_%s%s.png' % (OUT_DIR, output_name, cur_roi,
suffix), dpi=150)
# plt.savefig('%s/%s_%s.png' % (OUT_DIR, output_name, cur_roi), dpi=150)
except:
pass
# # average CCs by network (to stablize the estimates that we input into the BHM), HPinterc
# # net-GP
# In [22]: np.array(full_labels)[net_GP_labels > 0]
# Out[22]: array(['pGP-rh', 'aGP-rh', 'pGP-lh', 'aGP-lh'], dtype='<U36')
net_GP_labels = np.array([(s.find('GP') != -1) * 20 for s in full_labels])
for cur_net1 in np.unique(net17_labels):
for cur_net2 in np.unique(net_GP_labels):
if cur_net2 == 0:
continue
print('-' * 80)
cur_net_name1 = net17_label_names[cur_net1 - 1]
cur_net_name2 = 'GP'
print(cur_net_name1)
print('-- and --')
print(cur_net_name2)
net_inds1 = np.where(net17_labels == cur_net1)[0]
net_inds2 = np.where(net_GP_labels == cur_net2)[0]
y = CCs_all[:, net_inds1, :][:, :, net_inds2]
y = y.reshape((len(y), len(net_inds1)*len(net_inds2))).mean(1)
print(pd.DataFrame(y).describe())
# count 273.000000
# mean 0.340264
# std 0.134900
# min 0.078441
# 25% 0.238487
# 50% 0.314088
# 75% 0.428571
# max 0.750653
n_drugs = len(np.unique(drug_all))
pm_varnames = []
with pm.Model() as hierarchical_model:
hp_drugclass = pm.Normal('hp_drugclass', mu=0., sigma=1, shape=n_drugs)
hp_drugclass_priors = list(np.zeros(n_studies))
for j_drug in range(n_drugs):
for k_study in np.unique(study_id_all[drug_all == j_drug]):
hp_drugclass_priors[k_study] = hp_drugclass[j_drug]
alpha_drugclass = pm.Normal('alpha_drugclass', mu=hp_drugclass_priors, sigma=1, shape=n_studies)
# hp_cond = pm.Normal('hp_cond', mu=0., sigma=1, shape=1)
# alpha_cond = pm.Normal('alpha_cond', mu=hp_cond, sigma=1, shape=2)
# alpha_cond = pm.Normal('alpha_cond', mu=0, sigma=1, shape=2)
# beta_cond = pm.Normal('beta_cond', mu=0, sigma=1)
hp_beta = pm.Normal('hp_beta', mu=0., sigma=1, shape=1)
slope_betas = pm.Normal('slope_betas', mu=hp_beta, sigma=1, shape=n_drugs)
# beta_study_priors = list(np.zeros(n_studies))
# for j_drug in range(n_drugs):
# for k_study in np.unique(study_id_all[drug_all == j_drug]):
# beta_study_priors[k_study] = hp_beta[j_drug]
# beta_inter = pm.Normal('beta_inter_cond_drug', mu=beta_study_priors, sigma=2, shape=n_studies)
# reg_effect = alpha_cond[conds]
reg_effect = alpha_drugclass[study_id_all] + slope_betas[drug_all] * conds_all
eps = pm.HalfCauchy('eps', 5) # Model error
group_like = pm.Normal('beh_like', mu=reg_effect, sigma=eps, observed=y)
with hierarchical_model:
hierarchical_trace = pm.sample(draws=10000, n_init=5000, #init='advi',
chains=3, cores=9, progressbar=True,
random_seed=[1, 2, 3]) # one per chain needed
print(pm.summary(hierarchical_trace, hdi_prob=0.66))
OUT_DIR = '_results_uncorrected_data_pipelines_ana8'
output_name = 'net-GP_HPinterc_effect_%s-%s' % (cur_net_name1, cur_net_name2)
t = hierarchical_trace
for cur_roi in hierarchical_model.named_vars:
from matplotlib.lines import Line2D
cur_roi = cur_roi
plt.close('all')
# THRESH = 0.5
# THRESH = 1.0
n_last_chains = 500
try:
suffix = ''
fig = arviz.plot_posterior(t, var_names=cur_roi, textsize=15.0,
hdi_prob=0.66)
# fig = pm.plot_posterior(t[-n_last_chains:], varnames=[cur_roi],
#

ana8_uncorrected_data_pipelines_250820_NM.py at commit 0c58406, no license · at the source

Overview

Authors: Manesh Girn¹, Manoj K Doss², Leor Roseman³, Katrin H Preller⁴, Fernanda Palhano-Fontes⁵, Lorenzo Pasquini¹, Frederick S Barrett⁶, Pablo Mallaroni⁷, Natasha L Mason⁷, Christopher Timmermann³, Drummond E McCulloch⁸, Patrick M Fisher⁸, Brian S Winston⁶, Flora Moujaes⁴, Felix Muller⁹, Matthias E Liechti⁹, Franz X Vollenweider⁴, Johannes G Ramaekers⁷, Kim Kuypers⁷, Draulio B Araujo⁵

and 7 other authors

Olaf Sporns¹⁰, Joshua Siegel¹¹, Nico Dosenbach¹¹, David J Nutt³, Robin L Carhart-Harris¹, Emmanuel A Stamatakis¹², Danilo Bzdok^13,14

14 affiliations

Institutions: University of California, San Francisco (United States); The University of Texas at Austin (United States); University of Exeter (United Kingdom); University of Zurich (Switzerland); Universidade Federal do Rio Grande do Norte (Brazil); Johns Hopkins University (United States); Johns Hopkins Medicine (United States); Maastricht University (Netherlands); Rigshospitalet (Denmark); University of Basel (Switzerland); University Hospital of Basel (Switzerland); Indiana University Bloomington (United States); Indiana University (United States); NYU Langone Health (United States); New York University (United States); University of Cambridge (United Kingdom); Montreal Neurological Institute and Hospital (Canada); Mila - Quebec Artificial Intelligence Institute (Canada); McGill University (Canada)

Journal: Nature medicine, volume 32, issue 4, pages 1543-1554

Dates: received 25 March 2025; accepted 12 February 2026; published online 6 April 2026; in print 2026

Type: Research article · Language: English

License: CC BY-NC-ND

Identifiers: DOI 10.1038/s41591-026-04287-9 · PMID 41942645 · PMCID PMC13099416 · OpenAlex W7150978140

Open access: hybrid, a free copy (OpenAlex)

Status: code verified

Categories: fMRI (modality), human (organism), systems (subfield)

Methods: Connectivity, Smoothing, state filtering, decompositions, fMRI & imaging

Keywords: Computational biology and bioinformatics, Medical research

MeSH: Brain*, Hallucinogens*, Nerve Net*, Banisteriopsis, Bayes Theorem, Brain Mapping, Humans, Lysergic Acid Diethylamide, Magnetic Resonance Imaging, N,N-Dimethyltryptamine, Psilocybin (* major topic)

Topic: Psychedelics and Drug Studies (Clinical Psychology, Psychology), according to OpenAlex

Funding: NIA NIH HHS (T32 AG027668)

Citations: cited by 12 papers (Europe PMC); 67 references in the paper

Abstract

The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.

Repository

Its files are read in the Code ↔ Paper reader above, with 2 matches between paragraphs and lines of code.

banilo/BOLD_psychedelics_consortium

License: none: the authors keep all their rights

State: the link answers, verified on 29 September 2026

Evidence: files inventoried

Commit: 0c5840620e54026734f9035a6d49ad54ef46d9a4, 11 December 2025

Languages: Python (1)

Size: 12,692 files, 1 script

Software Heritage: not archived

Found in: “Code availability”

Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: ArviZ (1 file), Matplotlib (1 file), NiBabel (1 file), Nilearn (1 file), NumPy (1 file), pandas (1 file), PyMC (1 file), scikit-learn (1 file), SciPy (1 file), seaborn (1 file)

Availability: 1 check, the latest on 29 September 2026: the link answers

29 September 2026: the link answers

2 files

ana8_uncorrected_data_pipelines_250820_NM.py, Python, 4,756 lines, 2 matches
README.md, Text, 1 line

At the source: github.com/banilo/BOLD_psychedelics_consortium

Code availability statement

The paper has a code availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

it points to the authors' code: banilo/BOLD_psychedelics_consortium

Read it in the paper: doi.org/10.1038/s41591-026-04287-9.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
1 script, each with its path and the digest of its content;
2 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability statement

The paper has a data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

no repository, dataset or request procedure was recognized in it

Read it in the paper: doi.org/10.1038/s41591-026-04287-9.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 29 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 27 authors, 2 keywords, 11 MeSH terms, 1 funder, 64 references.

Cite

This paper

Girn, M., Doss, M. K., Roseman, L., Preller, K. H., Palhano-Fontes, F., Pasquini, L., Barrett, F. S., Mallaroni, P., Mason, N. L., Timmermann, C., McCulloch, D. E., Fisher, P. M., Winston, B. S., Moujaes, F., Muller, F., Liechti, M. E., Vollenweider, F. X., Ramaekers, J. G., Kuypers, K., . . . Bzdok, D. (2026). An international mega-analysis of psychedelic drug effects on brain circuit function. Nature medicine, 32(4), 1543-1554. https://doi.org/10.1038/s41591-026-04287-9

BibTeX

@article{girn2026international, author = {Girn, Manesh and Doss, Manoj K and Roseman, Leor and Preller, Katrin H and Palhano-Fontes, Fernanda and Pasquini, Lorenzo and Barrett, Frederick S and Mallaroni, Pablo and Mason, Natasha L and Timmermann, Christopher and McCulloch, Drummond E and Fisher, Patrick M and Winston, Brian S and Moujaes, Flora and Muller, Felix and Liechti, Matthias E and Vollenweider, Franz X and Ramaekers, Johannes G and Kuypers, Kim and Araujo, Draulio B and Sporns, Olaf and Siegel, Joshua and Dosenbach, Nico and Nutt, David J and Carhart-Harris, Robin L and Stamatakis, Emmanuel A and Bzdok, Danilo}, title = {{An international mega-analysis of psychedelic drug effects on brain circuit function}}, journal = {Nature medicine}, year = {2026}, month = apr, volume = {32}, number = {4}, pages = {1543--1554}, publisher = {Nature Portfolio}, issn = {1078-8956}, doi = {10.1038/s41591-026-04287-9}, url = {https://doi.org/10.1038/s41591-026-04287-9}, pmid = {41942645}, pmcid = {PMC13099416} }

RIS

TY - JOUR AU - Girn, Manesh AU - Doss, Manoj K AU - Roseman, Leor AU - Preller, Katrin H AU - Palhano-Fontes, Fernanda AU - Pasquini, Lorenzo AU - Barrett, Frederick S AU - Mallaroni, Pablo AU - Mason, Natasha L AU - Timmermann, Christopher AU - McCulloch, Drummond E AU - Fisher, Patrick M AU - Winston, Brian S AU - Moujaes, Flora AU - Muller, Felix AU - Liechti, Matthias E AU - Vollenweider, Franz X AU - Ramaekers, Johannes G AU - Kuypers, Kim AU - Araujo, Draulio B AU - Sporns, Olaf AU - Siegel, Joshua AU - Dosenbach, Nico AU - Nutt, David J AU - Carhart-Harris, Robin L AU - Stamatakis, Emmanuel A AU - Bzdok, Danilo TI - An international mega-analysis of psychedelic drug effects on brain circuit function T2 - Nature medicine J2 - Nat Med PY - 2026 DA - 2026/04/06 VL - 32 IS - 4 SP - 1543 EP - 1554 SN - 1078-8956 PB - Nature Portfolio DO - 10.1038/s41591-026-04287-9 UR - https://doi.org/10.1038/s41591-026-04287-9 LA - en ER -

CSL-JSON

{ "id": "10.1038/s41591-026-04287-9", "type": "article-journal", "title": "An international mega-analysis of psychedelic drug effects on brain circuit function", "container-title": "Nature medicine", "author": [ { "family": "Girn", "given": "Manesh" }, { "family": "Doss", "given": "Manoj K" }, { "family": "Roseman", "given": "Leor" }, { "family": "Preller", "given": "Katrin H" }, { "family": "Palhano-Fontes", "given": "Fernanda" }, { "family": "Pasquini", "given": "Lorenzo" }, { "family": "Barrett", "given": "Frederick S" }, { "family": "Mallaroni", "given": "Pablo" }, { "family": "Mason", "given": "Natasha L" }, { "family": "Timmermann", "given": "Christopher" }, { "family": "McCulloch", "given": "Drummond E" }, { "family": "Fisher", "given": "Patrick M" }, { "family": "Winston", "given": "Brian S" }, { "family": "Moujaes", "given": "Flora" }, { "family": "Muller", "given": "Felix" }, { "family": "Liechti", "given": "Matthias E" }, { "family": "Vollenweider", "given": "Franz X" }, { "family": "Ramaekers", "given": "Johannes G" }, { "family": "Kuypers", "given": "Kim" }, { "family": "Araujo", "given": "Draulio B" }, { "family": "Sporns", "given": "Olaf" }, { "family": "Siegel", "given": "Joshua" }, { "family": "Dosenbach", "given": "Nico" }, { "family": "Nutt", "given": "David J" }, { "family": "Carhart-Harris", "given": "Robin L" }, { "family": "Stamatakis", "given": "Emmanuel A" }, { "family": "Bzdok", "given": "Danilo" } ], "container-title-short": "Nat Med", "volume": "32", "issue": "4", "page": "1543-1554", "DOI": "10.1038/s41591-026-04287-9", "PMID": "41942645", "PMCID": "PMC13099416", "ISSN": "1078-8956", "publisher": "Nature Portfolio", "URL": "https://doi.org/10.1038/s41591-026-04287-9", "language": "en", "issued": { "date-parts": [ [ 2026, 4, 6 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.