OSCR

APOE ε4 as a predictor of cognitive decline and its interaction with hippocampal volume in Alzheimer's disease.

Code ↔ Paper

20 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 20 matches
  1. [1] § Materials and methods › Variables ↔ notebooks2/01_data_pipeline.ipynb, lines 288–360 · score 0.87 · ST120SV, ST29SV, ST30SV, FreeSurfer, ICV adjusted, HIPPO_ICV_ADJ
  2. [2] § Materials and methods › Statistical analysis › Linear mixed-effects models ↔ notebooks2/03_lme_analysis.ipynb, lines 279–343 · score 0.78 · likelihood ratio, APOE4 interaction terms, way interaction term, reduced model, nested, LRT
  3. [3] § Materials and methods › Variables ↔ notebooks2/02_descriptive_stats.ipynb, lines 198–248 · score 0.73 · 0–18, 0–85, 0–30, ADAS Cog13, CDR SB, worse
  4. [4] § Materials and methods › Statistical analysis › Survival analysis ↔ notebooks2/04_survival_sensitivity.ipynb, lines 119–162 · score 0.72 · Kaplan Meier curves, log rank, APOE4_DOSE, lifelines, fitted, Survival
  5. [5] § Materials and methods › Statistical analysis › Sensitivity analyses ↔ notebooks2/04_survival_sensitivity.ipynb, lines 381–421 · score 0.68 · selection bias, multi visit, single visit, LME, sensitivity, MMSE
  6. [6] § Materials and methods › Statistical analysis › Model assumptions ↔ notebooks2/03_lme_analysis.ipynb, lines 199–268 · score 0.66 · Shapiro Wilk, LME models, Residuals, SD, MMSE
  7. [7] § Materials and methods › Statistical analysis › Linear mixed-effects models ↔ notebooks2/04_survival_sensitivity.ipynb, lines 212–297 · score 0.65 · random intercept, random slope, HIPPO_ICV_ADJ, formulated, baseline diagnosis, DX
  8. [8] § Materials and methods › Statistical analysis › Descriptive statistics ↔ notebooks2/02_descriptive_stats.ipynb, lines 45–153 · score 0.65 · chi squared, Kruskal Wallis, Baseline characteristics, SD, variables, dose
  9. [9] § Materials and methods › Statistical analysis › Linear mixed-effects models ↔ notebooks2/03_lme_analysis.ipynb, lines 77–152 · score 0.62 · random intercepts, random slopes, statsmodels, LME, fitted
  10. [10] § Materials and methods › Statistical analysis › Linear mixed-effects models ↔ notebooks2/04_survival_sensitivity.ipynb, lines 212–297 · score 0.62 · random intercepts, random slopes, statsmodels, fitted, LME, visit
  11. [11] § Results › APOE ε4 × hippocampal volume × time interaction ↔ notebooks2/03_lme_analysis.ipynb, lines 279–343 · score 0.60 · LME model, way interaction term, APOE4_DOSE, HIPPO_ICV_ADJ, MMSE
  12. [12] § Results › Survival analysis: conversion risk ↔ notebooks2/04_survival_sensitivity.ipynb, lines 119–162 · score 0.60 · Kaplan Meier curves, log rank, CN MCI, Survival, event, conversion
  13. [13] § Materials and methods › Workflow overview ↔ notebooks2/04_survival_sensitivity.ipynb, lines 1–59 · score 0.60 · Kaplan Meier curves, robust standard errors, MMSE ceiling, survival, Cox, outliers
  14. [14] § Materials and methods › Preprocessing and quality control ↔ notebooks2/01_data_pipeline.ipynb, lines 288–360 · score 0.58 · FreeSurfer, ICV adjusted, scans, pipeline, Raw, VISCODE2
  15. [15] § Results › Baseline characteristics ↔ notebooks2/02_descriptive_stats.ipynb, lines 45–153 · score 0.57 · Kruskal Wallis, ADAS Cog13, CDR SB, baseline characteristics, NPI, SD
  16. [16] § Materials and methods › Statistical analysis › Linear mixed-effects models ↔ notebooks2/03_lme_analysis.ipynb, lines 77–152 · score 0.55 · random intercept, random slope, formulated, HIPPO_ICV_ADJ, DX, model
  17. [17] § Results › Survival analysis: conversion risk ↔ notebooks2/04_survival_sensitivity.ipynb, lines 164–210 · score 0.53 · Cox proportional hazards, AD conversion, survival, fitted, predictor, events
  18. [18] § Results › Survival analysis: conversion risk ↔ notebooks2/04_survival_sensitivity.ipynb, lines 164–210 · score 0.52 · hazard ratios, Cox model, Survival, conversion, interaction, dose
  19. [19] § Results › Sensitivity analyses ↔ notebooks2/04_survival_sensitivity.ipynb, lines 483–514 · score 0.51 · Cluster robust, standard errors, OLS, fitted, LME, Sensitivity
  20. [20] § Results › APOE ε4 × hippocampal volume × time interaction ↔ notebooks2/02_descriptive_stats.ipynb, lines 385–416 · score 0.51 · MMSE score, ICV adjusted hippocampal, Regression, hippocampal volume, correlation, baseline

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 authors' code

Jupyter notebook · 570 lines · 23 KB · MIT · 9 matches

  1. # %% [markdown]
  2. # # Survival Analysis & Sensitivity Analyses
  3. # **Study:** APOE ε4 × Hippocampal Volume × Cognitive Decline
  4. #
  5. # This notebook covers:
  6. # 1. **Survival/Conversion Analysis** — Cox PH model for CN → MCI/AD conversion
  7. # 2. **Kaplan-Meier Curves** by APOE dose
  8. # 3. **Sensitivity Analysis 1** — Stratified by baseline diagnosis (CN, MCI, AD)
  9. # 4. **Sensitivity Analysis 2** — Excluding outliers (±3 SD hippocampal volume)
  10. # 5. **Sensitivity Analysis 3** — Subjects with ≥2 visits only
  11. # 6. **Sensitivity Analysis 4** — Robust standard errors (MMSE ceiling effect)
  12. #
  13. # Requires: `reports/ADNI_Complete_Cases.csv` and `reports/ADNI_Baseline_Analysis.csv` from `01_data_pipeline.ipynb`
  15. # %%
  16. import pandas as pd
  17. import numpy as np
  18. import matplotlib.pyplot as plt
  19. import matplotlib.gridspec as gridspec
  20. import seaborn as sns
  21. import statsmodels.formula.api as smf
  22. import statsmodels.api as sm
  23. from scipy import stats
  24. import os
  25. import warnings
  26. warnings.filterwarnings('ignore')
  28. # Try to import lifelines for survival analysis
  29. try:
  30. from lifelines import CoxPHFitter, KaplanMeierFitter
  31. from lifelines.statistics import logrank_test, multivariate_logrank_test
  32. LIFELINES = True
  33. print('✓ lifelines available')
  34. except ImportError:
  35. LIFELINES = False
  36. print('⚠ lifelines not installed — install with: pip install lifelines')
  37. print(' Survival analysis will be skipped')
  39. BASE = '/media/faizaan/4TB/1_DATA_PROJECTS/Projects/Multimodel_study'
  40. REPORTS = os.path.join(BASE, 'reports')
  42. plt.rcParams.update({'figure.dpi': 120, 'font.size': 11,
  43. 'axes.spines.top': False, 'axes.spines.right': False})
  44. APOE_COLORS = {0: '#2196F3', 1: '#FF9800', 2: '#F44336'}
  46. # Load data
  47. complete = pd.read_csv(os.path.join(REPORTS, 'ADNI_Complete_Cases.csv'))
  48. baseline = pd.read_csv(os.path.join(REPORTS, 'ADNI_Baseline_Analysis.csv'))
  50. for df in [complete, baseline]:
  51. df['APOE4_DOSE'] = pd.to_numeric(df['APOE4_DOSE'], errors='coerce')
  52. df['HIPPO_ICV_ADJ'] = pd.to_numeric(df['HIPPO_ICV_ADJ'], errors='coerce')
  53. df['YEARS_FROM_BL'] = pd.to_numeric(df['YEARS_FROM_BL'], errors='coerce')
  54. df['MMSCORE'] = pd.to_numeric(df['MMSCORE'], errors='coerce')
  55. if 'SEX' in df.columns:
  56. df['SEX_MALE'] = (df['SEX'] == 'Male').astype(float)
  58. print(f'Complete: {complete["RID"].nunique()} subjects, {len(complete):,} obs')
  59. print(f'Baseline: {len(baseline)} subjects')
  61. # %% [markdown]
  62. # ## 1. Survival Analysis — CN → MCI/AD Conversion
  63. # Using pre-computed survival variables (`Event`, `Time`) from `baseline_subj.csv`.
  65. # %%
  66. # Prepare survival dataset
  67. # Keep CN subjects at baseline (those at risk of converting)
  68. surv_cols_needed = ['RID', 'APOE4_DOSE', 'HIPPO_ICV_ADJ', 'HIPPO_ICV_Z',
  69. 'AGE', 'SEX_MALE', 'PTEDUCAT', 'BL_DX_LABEL']
  70. event_cols = ['Event', 'Time', 'TimeMonths', 'TIME_YEARS']
  72. available_surv = [c for c in surv_cols_needed + event_cols if c in baseline.columns]
  73. surv_df = baseline[available_surv].copy()
  75. # Check survival columns
  76. has_event = 'Event' in surv_df.columns
  77. has_time = 'TIME_YEARS' in surv_df.columns
  79. print(f'Survival columns available: Event={has_event}, Time={has_time}')
  81. if has_event and has_time:
  82. # TIME_YEARS already converted in NB01 (was in months, now years)
  83. surv_df = surv_df.dropna(subset=['Event', 'TIME_YEARS', 'APOE4_DOSE', 'HIPPO_ICV_ADJ'])
  84. surv_df['TIME_YEARS'] = surv_df['TIME_YEARS'].clip(lower=0.01) # avoid zero-time
  86. print(f'\nSurvival dataset: {len(surv_df)} subjects')
  87. print(f'Events (conversions): {surv_df["Event"].sum():.0f} ({surv_df["Event"].mean()*100:.1f}%)')
  88. print(f'Follow-up (years): median={surv_df["TIME_YEARS"].median():.1f}, max={surv_df["TIME_YEARS"].max():.1f}')
  89. print('\nEvents by APOE dose:')
  90. print(surv_df.groupby('APOE4_DOSE')[['Event']].agg(['sum', 'count']))
  91. else:
  92. print('\n⚠ Survival variables (Event, Time) not found in baseline data.')
  93. print(' Will derive conversion status from longitudinal diagnosis data.')
  95. # Derive from longitudinal data
  96. if 'BL_DX_LABEL' in complete.columns and 'DX_LABEL' in complete.columns:
  97. cn_subjects = complete[complete['BL_DX_LABEL'] == 'CN']['RID'].unique()
  99. conversions = []
  100. for rid in cn_subjects:
  101. sub = complete[complete['RID'] == rid].sort_values('YEARS_FROM_BL')
  102. converted = (sub['DX_LABEL'].isin(['MCI', 'AD'])).any()
  103. if converted:
  104. conv_time = sub[sub['DX_LABEL'].isin(['MCI', 'AD'])]['YEARS_FROM_BL'].min()
  105. else:
  106. conv_time = sub['YEARS_FROM_BL'].max()
  107. conversions.append({'RID': rid, 'Event': int(converted), 'Time': conv_time})
  109. conv_df = pd.DataFrame(conversions)
  110. surv_df = baseline[baseline['RID'].isin(cn_subjects)].merge(conv_df, on='RID', how='inner')
  111. surv_df = surv_df.dropna(subset=['Event', 'Time', 'APOE4_DOSE', 'HIPPO_ICV_ADJ'])
  112. has_event = True
  113. has_time = True
  114. print(f'Derived conversion: {len(surv_df)} CN subjects, {surv_df["Event"].sum():.0f} conversions')
  115. else:
  116. print('Cannot derive conversion status — no DX_LABEL in longitudinal data')
  117. has_event = False
  119. # %% [markdown]
  120. # ### 1.1 Kaplan-Meier Curves by APOE Dose
  122. # %%
  123. if LIFELINES and has_event and 'surv_df' in locals() and len(surv_df) > 10:
  124. fig, ax = plt.subplots(figsize=(10, 6))
  126. kmfs = {}
  127. for dose in [0, 1, 2]:
  128. sub = surv_df[surv_df['APOE4_DOSE'] == dose]
  129. if len(sub) < 5: continue
  130. kmf = KaplanMeierFitter()
  131. kmf.fit(sub['TIME_YEARS'], sub['Event'], label=f'{dose} ε4 alleles (n={len(sub)})')
  132. kmf.plot_survival_function(ax=ax, color=APOE_COLORS[dose], ci_show=True, ci_alpha=0.15)
  133. kmfs[dose] = kmf
  135. ax.set_xlabel('Time (years)', fontsize=12)
  136. ax.set_ylabel('Conversion-free survival probability', fontsize=12)
  137. ax.set_title('Kaplan-Meier Survival Curves\nCN → MCI/AD Conversion by APOE ε4 Dose', fontweight='bold')
  138. ax.set_ylim(0, 1.05)
  139. ax.legend(title='APOE ε4 dose')
  141. # Log-rank test
  142. if len(kmfs) >= 2:
  143. try:
  144. from lifelines.statistics import multivariate_logrank_test
  145. results_lr = multivariate_logrank_test(surv_df['TIME_YEARS'], surv_df['APOE4_DOSE'], surv_df['Event'])
  146. p_lr = results_lr.p_value
  147. ax.text(0.98, 0.97, f'Log-rank p={"<0.001" if p_lr<0.001 else f"{p_lr:.3f}"}',
  148. transform=ax.transAxes, ha='right', va='top', fontsize=10,
  149. bbox=dict(boxstyle='round', facecolor='white', alpha=0.8))
  150. print(f'Log-rank test: p={p_lr:.4f}')
  151. except Exception as e:
  152. print(f'Log-rank test failed: {e}')
  154. plt.tight_layout()
  155. plt.savefig(os.path.join(REPORTS, 'Fig7_KaplanMeier.png'), dpi=150, bbox_inches='tight')
  156. plt.show()
  157. print('✓ Saved: Fig7_KaplanMeier.png')
  159. elif not LIFELINES:
  160. print('lifelines not available — install with: pip install lifelines')
  161. else:
  162. print('Insufficient survival data for KM curves')
  164. # %% [markdown]
  165. # ### 1.2 Cox Proportional Hazards Model
  167. # %%
  168. if LIFELINES and has_event and 'surv_df' in locals() and len(surv_df) > 10:
  169. # Build Cox model predictors
  170. cox_vars = ['APOE4_DOSE', 'HIPPO_ICV_ADJ']
  171. for v in ['AGE', 'SEX_MALE', 'PTEDUCAT', 'GDTOTAL']:
  172. if v in surv_df.columns:
  173. cox_vars.append(v)
  175. # Add APOE×Hippo interaction
  176. surv_df['APOE4_x_HIPPO'] = surv_df['APOE4_DOSE'] * surv_df['HIPPO_ICV_ADJ']
  177. cox_vars.append('APOE4_x_HIPPO')
  179. cox_cols = ['TIME_YEARS', 'Event'] + cox_vars
  180. cox_data = surv_df[[c for c in cox_cols if c in surv_df.columns]].dropna().copy()
  182. print(f'Cox model data: {len(cox_data)} subjects, {cox_data["Event"].sum():.0f} events')
  183. print(f'Predictors: {[c for c in cox_vars if c in cox_data.columns]}')
  185. cph = CoxPHFitter(penalizer=0.1) # small regularization for stability
  186. try:
  187. cph.fit(cox_data, duration_col='TIME_YEARS', event_col='Event')
  188. cph.print_summary()
  190. # Save Cox results
  191. cox_summary = cph.summary
  192. cox_summary.to_csv(os.path.join(REPORTS, 'Cox_PH_Results.csv'))
  193. print('\n✓ Saved: Cox_PH_Results.csv')
  195. # Forest plot of hazard ratios
  196. fig, ax = plt.subplots(figsize=(9, 5))
  197. cph.plot(ax=ax)
  198. ax.set_title('Cox PH Model — Hazard Ratios (95% CI)\nOutcome: CN → MCI/AD Conversion', fontweight='bold')
  199. plt.tight_layout()
  200. plt.savefig(os.path.join(REPORTS, 'Fig8_Cox_HazardRatios.png'), dpi=150, bbox_inches='tight')
  201. plt.show()
  202. print('✓ Saved: Fig8_Cox_HazardRatios.png')
  204. except Exception as e:
  205. print(f'Cox model failed: {e}')
  207. elif not LIFELINES:
  208. print('lifelines not available — skipping Cox model')
  209. else:
  210. print('Insufficient data for Cox model')
  212. # %% [markdown]
  213. # ## 2. Sensitivity Analysis 1 — Stratified by Baseline Diagnosis
  214. # Run LME models separately for CN, MCI, and AD subgroups.
  216. # %%
  217. def build_covariates(df):
  218. cov_parts = []
  219. if 'AGE' in df.columns: cov_parts.append('AGE')
  220. if 'SEX_MALE' in df.columns: cov_parts.append('SEX_MALE')
  221. if 'PTEDUCAT' in df.columns: cov_parts.append('PTEDUCAT')
  222. if 'GDTOTAL' in df.columns: cov_parts.append('GDTOTAL')
  223. return ' + '.join(cov_parts)
  225. def run_lme_simple(data, outcome, cov_str, label):
  226. """Fit LME with random intercept only (robust to small samples)"""
  227. if outcome not in data.columns:
  228. return None
  229. data_clean = data.dropna(subset=[outcome, 'YEARS_FROM_BL', 'APOE4_DOSE', 'HIPPO_ICV_ADJ']).copy()
  230. if 'GDTOTAL' in data_clean.columns:
  231. data_clean['GDTOTAL'] = data_clean['GDTOTAL'].fillna(0)
  233. # Need ≥2 visits per subject
  234. vc = data_clean.groupby('RID').size()
  235. data_clean = data_clean[data_clean['RID'].isin(vc[vc>=2].index)]
  237. n_sub = data_clean['RID'].nunique()
  238. if n_sub < 10:
  239. print(f' {label}: only {n_sub} subjects — too few for LME, skipping')
  240. return None
  242. formula = f"{outcome} ~ YEARS_FROM_BL * APOE4_DOSE * HIPPO_ICV_ADJ" + \
  243. (f" + {cov_str}" if cov_str else "")
  245. # Align data with Patsy's NaN-dropping to prevent statsmodels index mismatch
  246. import patsy as _patsy
  247. try:
  248. _, _X = _patsy.dmatrices(formula, data=data_clean, return_type='dataframe')
  249. data_clean = data_clean.loc[_X.index].reset_index(drop=True)
  250. n_sub = data_clean['RID'].nunique()
  251. except Exception:
  252. data_clean = data_clean.reset_index(drop=True)
  254. try:
  255. # Try random slope first
  256. m = smf.mixedlm(formula, data=data_clean, groups=data_clean['RID'],
  257. re_formula='~YEARS_FROM_BL')
  258. r = m.fit(method='lbfgs', maxiter=300, disp=False)
  259. if not np.isnan(r.llf):
  260. print(f' {label}: n={n_sub}, random slope converged (AIC={r.aic:.1f})')
  261. return r
  262. except:
  263. pass
  265. try:
  266. m = smf.mixedlm(formula, data=data_clean, groups=data_clean['RID'])
  267. r = m.fit(method='lbfgs', maxiter=300, disp=False)
  268. print(f' {label}: n={n_sub}, random intercept only (AIC={r.aic:.1f})')
  269. return r
  270. except Exception as e:
  271. print(f' {label}: FAILED — {str(e)[:60]}')
  272. return None
  274. print('=== SENSITIVITY 1: Stratified by Baseline Diagnosis ===')
  275. print('MMSE models by BL diagnosis:\n')
  277. strat_results = {}
  278. for dx in ['CN', 'MCI', 'AD']:
  279. if 'BL_DX_LABEL' not in complete.columns:
  280. print(f' BL_DX_LABEL not available — cannot stratify')
  281. break
  282. sub = complete[complete['BL_DX_LABEL'] == dx].copy()
  283. cov = build_covariates(sub)
  284. r = run_lme_simple(sub, 'MMSCORE', cov, f'MMSE-{dx}')
  285. strat_results[dx] = r
  287. print('\nResults summary:')
  288. for dx, r in strat_results.items():
  289. if r is not None:
  290. # Focus on key interaction terms
  291. key_terms = ['YEARS_FROM_BL', 'APOE4_DOSE', 'HIPPO_ICV_ADJ',
  292. 'YEARS_FROM_BL:APOE4_DOSE', 'YEARS_FROM_BL:HIPPO_ICV_ADJ']
  293. print(f'\n {dx} stratum:')
  294. for t in key_terms:
  295. if t in r.fe_params:
  296. p = r.pvalues[t]
  297. print(f' {t:<35} β={r.fe_params[t]:>7.3f}, p={"<0.001" if p<0.001 else f"{p:.3f}"}')
  299. # %%
  300. # Compare interaction coefficients across diagnostic groups (forest-style comparison)
  301. if any(v is not None for v in strat_results.values()):
  302. target_term = 'YEARS_FROM_BL:APOE4_DOSE'
  303. alt_terms = ['YEARS_FROM_BL:APOE4_DOSE', 'YEARS_FROM_BL:HIPPO_ICV_ADJ', 'APOE4_DOSE']
  305. comparison_rows = []
  306. for dx, r in strat_results.items():
  307. if r is None: continue
  308. for term in alt_terms:
  309. if term in r.fe_params:
  310. comparison_rows.append({
  311. 'Diagnosis': dx, 'Term': term,
  312. 'Coef': r.fe_params[term],
  313. 'CI_lo': r.conf_int().loc[term, 0],
  314. 'CI_hi': r.conf_int().loc[term, 1],
  315. 'P_value': r.pvalues[term]
  316. })
  318. if comparison_rows:
  319. comp_df = pd.DataFrame(comparison_rows)
  320. comp_df.to_csv(os.path.join(REPORTS, 'Sensitivity1_Stratified_Dx.csv'), index=False)
  321. print('\n✓ Saved: Sensitivity1_Stratified_Dx.csv')
  323. # Plot comparison
  324. for term in alt_terms:
  325. sub = comp_df[comp_df['Term'] == term]
  326. if len(sub) < 2: continue
  328. fig, ax = plt.subplots(figsize=(7, 4))
  329. colors_dx = {'CN': '#4CAF50', 'MCI': '#FF9800', 'AD': '#F44336'}
  330. for _, row in sub.iterrows():
  331. dx = row['Diagnosis']
  332. ax.errorbar(row['Coef'], dx,
  333. xerr=[[row['Coef']-row['CI_lo']], [row['CI_hi']-row['Coef']]],
  334. fmt='o', color=colors_dx.get(dx, 'grey'), capsize=5, markersize=8)
  335. ax.axvline(0, color='black', linewidth=1, linestyle='--', alpha=0.7)
  336. ax.set_xlabel('Coefficient (95% CI)')
  337. ax.set_title(f'Stratified Analysis: {term}\nMMSE Outcome', fontweight='bold')
  338. plt.tight_layout()
  339. safe_term = term.replace(':', '_').replace('/', '_')
  340. plt.savefig(os.path.join(REPORTS, f'FigStrat_{safe_term}.png'), dpi=150, bbox_inches='tight')
  341. plt.show()
  342. break # just first term
  344. # %% [markdown]
  345. # ## 3. Sensitivity Analysis 2 — Exclude Hippocampal Outliers (±3 SD)
  347. # %%
  348. print('=== SENSITIVITY 2: Exclude Hippocampal Outliers (±3 SD) ===')
  350. hippo_mean = complete['HIPPO_ICV_ADJ'].mean()
  351. hippo_sd = complete['HIPPO_ICV_ADJ'].std()
  352. low_cut = hippo_mean - 3*hippo_sd
  353. high_cut = hippo_mean + 3*hippo_sd
  355. print(f'Hippocampal mean: {hippo_mean:.4f}, SD: {hippo_sd:.4f}')
  356. print(f'Exclusion range: < {low_cut:.4f} or > {high_cut:.4f}')
  358. outlier_mask = (complete['HIPPO_ICV_ADJ'] < low_cut) | (complete['HIPPO_ICV_ADJ'] > high_cut)
  359. n_outliers = outlier_mask.sum()
  360. print(f'Outliers: {n_outliers} observations ({n_outliers/len(complete)*100:.1f}%)')
  362. no_outlier_df = complete[~outlier_mask].copy()
  363. print(f'\nPost-exclusion: {no_outlier_df["RID"].nunique()} subjects, {len(no_outlier_df):,} obs')
  365. cov_no = build_covariates(no_outlier_df)
  366. if 'GDTOTAL' in no_outlier_df.columns:
  367. no_outlier_df['GDTOTAL'] = no_outlier_df['GDTOTAL'].fillna(0)
  369. r_no_outlier = run_lme_simple(no_outlier_df, 'MMSCORE', cov_no, 'No-outlier MMSE')
  371. if r_no_outlier is not None:
  372. print('\nKey terms (no-outlier model):')
  373. key_terms = ['YEARS_FROM_BL', 'APOE4_DOSE', 'HIPPO_ICV_ADJ',
  374. 'YEARS_FROM_BL:APOE4_DOSE', 'YEARS_FROM_BL:HIPPO_ICV_ADJ',
  375. 'YEARS_FROM_BL:APOE4_DOSE:HIPPO_ICV_ADJ']
  376. for t in key_terms:
  377. if t in r_no_outlier.fe_params:
  378. p = r_no_outlier.pvalues[t]
  379. print(f' {t:<40} β={r_no_outlier.fe_params[t]:>8.4f}, p={"<0.001" if p<0.001 else f"{p:.3f}"}')
  381. # %% [markdown]
  382. # ## 4. Sensitivity Analysis 3 — Subjects with ≥2 Visits Only
  384. # %%
  385. print('=== SENSITIVITY 3: ≥2 Visits Only ===')
  387. visit_counts = complete.groupby('RID').size()
  388. multi_rids = visit_counts[visit_counts >= 2].index
  389. multi_df = complete[complete['RID'].isin(multi_rids)].copy()
  391. print(f'Original: {complete["RID"].nunique()} subjects')
  392. print(f'≥2 visits: {multi_df["RID"].nunique()} subjects ({multi_df["RID"].nunique()/complete["RID"].nunique()*100:.1f}%)')
  393. print(f'Mean visits: {visit_counts[multi_rids].mean():.1f}')
  395. # Compare demographics between single vs multi-visit (check selection bias)
  396. single_rids = visit_counts[visit_counts == 1].index
  397. if len(single_rids) > 0:
  398. bl_multi = baseline[baseline['RID'].isin(multi_rids)]
  399. bl_single = baseline[baseline['RID'].isin(single_rids)]
  401. print('\nSelection check — Multi-visit vs Single-visit subjects:')
  402. for var in ['AGE', 'MMSCORE', 'APOE4_DOSE', 'HIPPO_ICV_ADJ']:
  403. if var in bl_multi.columns:
  404. m_m = bl_multi[var].mean()
  405. m_s = bl_single[var].mean()
  406. _, p = stats.ttest_ind(bl_multi[var].dropna(), bl_single[var].dropna())
  407. print(f' {var:<18} Multi: {m_m:.2f} Single: {m_s:.2f} p={"<0.001" if p<0.001 else f"{p:.3f}"}')
  409. cov_multi = build_covariates(multi_df)
  410. if 'GDTOTAL' in multi_df.columns:
  411. multi_df['GDTOTAL'] = multi_df['GDTOTAL'].fillna(0)
  413. r_multi = run_lme_simple(multi_df, 'MMSCORE', cov_multi, '≥2-visit MMSE')
  415. if r_multi is not None:
  416. print('\nKey terms (≥2 visits model):')
  417. for t in ['YEARS_FROM_BL:APOE4_DOSE', 'YEARS_FROM_BL:HIPPO_ICV_ADJ',
  418. 'YEARS_FROM_BL:APOE4_DOSE:HIPPO_ICV_ADJ']:
  419. if t in r_multi.fe_params:
  420. p = r_multi.pvalues[t]
  421. print(f' {t:<45} β={r_multi.fe_params[t]:>8.4f}, p={"<0.001" if p<0.001 else f"{p:.3f}"}')
  423. # %% [markdown]
  424. # ## 5. Sensitivity Analysis 4 — Robust Standard Errors
  425. # MMSE has ceiling effects in CN subjects. Use robust SE to account for non-normality (Reviewer 2.3).
  427. # %%
  428. print('=== SENSITIVITY 4: MMSE Ceiling Effect & Robust Errors ===')
  430. # Show MMSE distribution
  431. fig, axes = plt.subplots(1, 3, figsize=(15, 4))
  433. # Overall MMSE distribution at baseline
  434. ax = axes[0]
  435. bl_mmse = baseline['MMSCORE'].dropna()
  436. ax.hist(bl_mmse, bins=20, color='steelblue', edgecolor='white', density=True)
  437. ax.axvline(28, color='red', linestyle='--', alpha=0.7, label='Ceiling zone (≥28)')
  438. ax.set_xlabel('MMSE score')
  439. ax.set_ylabel('Density')
  440. ax.set_title('MMSE Distribution (Baseline)\nNote ceiling effect near 30', fontweight='bold')
  441. ax.legend()
  442. ceil_pct = (bl_mmse >= 28).mean() * 100
  443. ax.text(0.05, 0.95, f'{ceil_pct:.1f}% ≥ 28', transform=ax.transAxes, va='top')
  445. # MMSE by APOE dose at baseline
  446. ax = axes[1]
  447. for dose in [0, 1, 2]:
  448. sub = baseline[baseline['APOE4_DOSE'] == dose]['MMSCORE'].dropna()
  449. ax.hist(sub, bins=15, alpha=0.5, color=APOE_COLORS[dose],
  450. label=f'APOE{dose} (n={len(sub)})', edgecolor='none', density=True)
  451. ax.set_xlabel('MMSE score')
  452. ax.set_ylabel('Density')
  453. ax.set_title('MMSE by APOE Dose (Baseline)', fontweight='bold')
  454. ax.legend(fontsize=9)
  456. # MMSE × Hippocampus scatter (all visits)
  457. ax = axes[2]
  458. if 'HIPPO_ICV_ADJ' in complete.columns:
  459. sample = complete.dropna(subset=['MMSCORE', 'HIPPO_ICV_ADJ']).sample(min(1000, len(complete)))
  460. ax.scatter(sample['HIPPO_ICV_ADJ'], sample['MMSCORE'],
  461. alpha=0.2, s=8, color='steelblue')
  462. slope, intercept, r, p, _ = stats.linregress(sample['HIPPO_ICV_ADJ'], sample['MMSCORE'])
  463. x = np.linspace(sample['HIPPO_ICV_ADJ'].min(), sample['HIPPO_ICV_ADJ'].max(), 100)
  464. ax.plot(x, slope*x + intercept, color='red', linewidth=2)
  465. ax.set_xlabel('ICV-adjusted hippocampal volume')
  466. ax.set_ylabel('MMSE score')
  467. ax.set_title(f'MMSE vs Hippocampus\nr={r:.2f}, p={"<0.001" if p<0.001 else f"{p:.3f}"}', fontweight='bold')
  469. plt.tight_layout()
  470. plt.savefig(os.path.join(REPORTS, 'FigSensitivity_MMSE_Distribution.png'), dpi=150, bbox_inches='tight')
  471. plt.show()
  472. print('✓ Saved: FigSensitivity_MMSE_Distribution.png')
  474. # Apply floor/ceiling sensitivity analysis
  475. print(f'\nMMSE ceiling analysis:')
  476. print(f' Overall: {(bl_mmse >= 28).mean()*100:.1f}% at ceiling (≥28)')
  477. if 'BL_DX_LABEL' in baseline.columns:
  478. for dx in ['CN', 'MCI', 'AD']:
  479. sub = baseline[baseline['BL_DX_LABEL'] == dx]['MMSCORE'].dropna()
  480. if len(sub) > 0:
  481. print(f' {dx}: {(sub >= 28).mean()*100:.1f}% at ceiling')
  483. # %%
  484. # Robust standard errors via statsmodels OLS with cluster-robust SE
  485. # (as sensitivity to LME)
  486. print('\nRobust SE approach — pooled OLS with cluster-robust SE (sensitivity check):')
  488. robust_data = complete.dropna(subset=['MMSCORE', 'YEARS_FROM_BL', 'APOE4_DOSE', 'HIPPO_ICV_ADJ']).copy()
  489. if 'GDTOTAL' in robust_data.columns:
  490. robust_data['GDTOTAL'] = robust_data['GDTOTAL'].fillna(0)
  492. cov_parts = [v for v in ['AGE', 'SEX_MALE', 'PTEDUCAT', 'GDTOTAL'] if v in robust_data.columns]
  493. if 'BL_DX_LABEL' in robust_data.columns:
  494. cov_parts.append('C(BL_DX_LABEL)')
  496. formula_ols = ("MMSCORE ~ YEARS_FROM_BL * APOE4_DOSE * HIPPO_ICV_ADJ" +
  497. (f" + {' + '.join(cov_parts)}" if cov_parts else ""))
  499. try:
  500. import patsy
  501. y, X = patsy.dmatrices(formula_ols, data=robust_data, return_type='dataframe')
  502. # Align groups to patsy's post-NaN-drop index
  503. robust_aligned = robust_data.loc[y.index].reset_index(drop=True)
  504. ols_model = sm.OLS(y, X)
  505. # Cluster-robust SE (clusters = subjects)
  506. ols_result_robust = ols_model.fit(cov_type='cluster', cov_kwds={'groups': robust_aligned['RID'].values})
  507. print(ols_result_robust.summary().tables[1])
  509. # Save
  510. with open(os.path.join(REPORTS, 'Sensitivity4_Robust_SE_MMSE.txt'), 'w') as f:
  511. f.write(ols_result_robust.summary().as_text())
  512. print('\n✓ Saved: Sensitivity4_Robust_SE_MMSE.txt')
  513. except Exception as e:
  514. print(f'Robust SE failed: {e}')
  516. # %% [markdown]
  517. # ## 6. Sensitivity Results Summary Table
  519. # %%
  520. # Compile key interaction terms across sensitivity analyses
  521. from importlib import import_module
  523. print('=== SENSITIVITY ANALYSIS SUMMARY ===')
  524. print('Key interaction term: Time × APOE4_DOSE × HIPPO_ICV_ADJ (3-way)\n')
  526. def extract_term(result, term):
  527. if result is None or term not in result.fe_params: return None, None, None
  528. coef = result.fe_params[term]
  529. ci_lo = result.conf_int().loc[term, 0]
  530. ci_hi = result.conf_int().loc[term, 1]
  531. p = result.pvalues[term]
  532. return coef, (ci_lo, ci_hi), p
  534. summary_rows = []
  535. for label, res in [
  536. ('No outliers (±3 SD)', r_no_outlier),
  537. ('≥2 visits only', r_multi),
  538. ('CN subgroup', strat_results.get('CN')),
  539. ('MCI subgroup', strat_results.get('MCI')),
  540. ('AD subgroup', strat_results.get('AD')),
  541. ]:
  542. for term in ['YEARS_FROM_BL:APOE4_DOSE', 'YEARS_FROM_BL:HIPPO_ICV_ADJ']:
  543. coef, ci, p = extract_term(res, term)
  544. if coef is not None:
  545. summary_rows.append({
  546. 'Analysis': label, 'Term': term,
  547. 'Coef': round(coef, 4),
  548. 'CI_95': f'[{ci[0]:.3f}, {ci[1]:.3f}]',
  549. 'P_value': '<0.001' if p < 0.001 else f'{p:.3f}',
  550. 'Sig': '***' if p<0.001 else '**' if p<0.01 else '*' if p<0.05 else ''
  551. })
  553. if summary_rows:
  554. sens_summary = pd.DataFrame(summary_rows)
  555. print(sens_summary.to_string(index=False))
  556. sens_summary.to_csv(os.path.join(REPORTS, 'Sensitivity_Summary_All.csv'), index=False)
  557. print('\n✓ Saved: Sensitivity_Summary_All.csv')
  558. else:
  559. print('No sensitivity results available to compile')
  561. # %% [markdown]
  562. # ## 7. Output Summary
  564. # %%
  565. print('SURVIVAL & SENSITIVITY ANALYSIS COMPLETE')
  566. print('='*60)
  567. print('\nFiles in reports/:')
  568. for f in sorted(os.listdir(REPORTS)):
  569. size = os.path.getsize(os.path.join(REPORTS, f)) / 1024
  570. print(f' {f:<50} {size:>8.1f} KB')

04_survival_sensitivity.ipynb at commit dc0e2c2, under MIT · at the source

Overview

Authors: Faizaan Fazal Khan1, Jun-Hyung Kim1, Ji-In Kim1, Goo-Rak Kwon1
  1. Department of Information and Communication Engineering, Chosun University, Gwangju, Republic of Korea
Institutions: Chosun University (South Korea)
Journal: Frontiers in aging neuroscience, volume 18, article 1730265
Dates: received 22 October 2025; accepted 6 April 2026; published online 22 April 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.3389/fnagi.2026.1730265 · PMID 42100481 · PMCID PMC13144130 · OpenAlex W7155159290
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: genetics / omics (modality), structural MRI / diffusion (modality), human (organism), Alzheimer's / dementia (population), clinical / translational (subfield)
Methods: Statistics
Keywords: Alzheimer’s disease, APOE ε4, cognitive decline, hippocampal atrophy, longitudinal trajectories, mixed-effects models, neuroimaging biomarkers, risk stratification
Topic: Dementia and Cognitive Impairment Research (Psychiatry and Mental health, Medicine), according to OpenAlex
Funding: Alzheimer's Disease Neuroimaging Initiative; National Research Foundation of Korea; National Institutes of Health; Canadian Institutes of Health Research
Citations: cited by 2 papers (Europe PMC); 15 references in the paper

Abstract

Background/Introduction: The apolipoprotein E (APOE) ε4 allele is the strongest known genetic risk factor for late-onset Alzheimer’s disease, and hippocampal atrophy is among the most reliable structural biomarkers of neurodegeneration. While both are independently associated with cognitive decline, whether APOE ε4 dose modulates the hippocampal volume–cognition relationship longitudinally in sporadic Alzheimer’s disease remains underexplored at adequate statistical power.

Methods: This study analyzed data from 2,417 Alzheimer’s Disease Neuroimaging Initiative participants with complete APOE genotypes, intracranial volume-adjusted hippocampal volumes, and longitudinal cognitive assessments spanning a mean follow-up of 4.2 years and up to 19.3 years with an average of 4.9 visits per participant. Linear mixed-effects models with random intercepts and slopes per subject estimated cognitive trajectories across the Mini-Mental State Examination (MMSE), Clinical Dementia Rating Sum of Boxes (CDR-SB), and Alzheimer’s Disease Assessment Scale Cognitive Subscale 13 (ADAS-Cog13) as a function of time, APOE ε4 dose, and ICV-adjusted hippocampal volume, including their three-way interaction and adjusting for age, sex, education, baseline diagnosis, and depression. Cox proportional hazards models were used to assess conversion risk.

Results: A clear APOE ε4 dose–response gradient was observed at baseline across all cognitive and hippocampal measures (all p < 0.001). Linear mixed-effects models revealed a significant three-way interaction of time × APOE ε4 dose × hippocampal volume on MMSE (β = −0.79, 95% CI [−1.51, −0.08], p = 0.030) and CDR-SB (β = +0.47, 95% CI [+0.03, +0.91], p = 0.037) trajectories, both significant under Bonferroni correction (α = 0.017), indicating that APOE ε4 amplifies the association between smaller hippocampal volume and faster cognitive deterioration over time. The time × hippocampal volume interaction was confirmed as highly significant by likelihood ratio test (LR = 3712.99, p < 0.001). Cox proportional hazards analyses of 845 conversion events showed that each additional ε4 allele conferred a 48% increase in conversion risk (HR = 1.48, 95% CI [1.29, 1.71], p < 0.001). Sensitivity analyses across diagnostic strata, after outlier exclusion, and in multi-visit subsamples confirmed the robustness of hippocampal volume effects.

Discussion/Conclusion: These findings demonstrate that APOE ε4 genotype significantly modulates the longitudinal relationship between hippocampal volume and cognitive decline, supporting the integration of APOE genotype and structural hippocampal imaging for refined individual risk stratification in Alzheimer’s disease.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 20 matches between paragraphs and lines of code.

FaizaanFazal/adni-apoe4-hippocampus-cognitive-decline

License: MIT
State: the link answers, verified on 29 September 2026
Evidence: files inventoried
Commit: dc0e2c2273170867e2b47f128e6bf30cc42e9b0c, 22 April 2026
Languages: Jupyter (5), Python (1)
Size: 40 files, 6 scripts
Software Heritage: not archived
Found in: “Data availability statement”
Holds: README, license file, environment (requirements_notebooks.txt), 5 notebooks
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (6 files), pandas (6 files), Matplotlib (3 files), SciPy (3 files), seaborn (3 files), statsmodels (2 files)
Availability: 1 check, the latest on 29 September 2026: the link answers
  • 29 September 2026: the link answers
8 files
At the source: github.com/FaizaanFazal/adni-apoe4-hippocampus-cognitive-decline

The paper's code and data availability statement is in the Data section.

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;
  • 6 scripts, each with its path and the digest of its content;
  • 20 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 and code availability

The data used in this study were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu) and are available to qualified researchers upon request. The analysis code supporting the findings of this study will be made available by the authors upon acceptance of the manuscript.

Reproduced under the paper's license (CC BY), from the paper cited above.

Data availability statement

The data used in this study were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database () and are available to qualified researchers upon application to ADNI. The analysis code supporting the findings of this study is publicly available at: https://github.com/FaizaanFazal/adni-apoe4-hippocampus-cognitive-decline.

Reproduced under the paper's license (CC BY), from the paper cited above.

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, pages, dates, 4 authors, 8 keywords, 4 funders, 15 references.

Cite

This paper

Khan, F. F., Kim, J.-H., Kim, J.-I., & Kwon, G.-R. (2026). APOE ε4 as a predictor of cognitive decline and its interaction with hippocampal volume in Alzheimer's disease. Frontiers in aging neuroscience, 18, 1730265. https://doi.org/10.3389/fnagi.2026.1730265

BibTeX

@article{khan2026apoe,
author = {Khan, Faizaan Fazal and Kim, Jun-Hyung and Kim, Ji-In and Kwon, Goo-Rak},
title = {{APOE ε4 as a predictor of cognitive decline and its interaction with hippocampal volume in Alzheimer's disease}},
journal = {Frontiers in aging neuroscience},
year = {2026},
month = apr,
volume = {18},
pages = {1730265},
publisher = {Frontiers Media SA},
issn = {1663-4365},
doi = {10.3389/fnagi.2026.1730265},
url = {https://doi.org/10.3389/fnagi.2026.1730265},
pmid = {42100481},
pmcid = {PMC13144130}
}

RIS

TY - JOUR
AU - Khan, Faizaan Fazal
AU - Kim, Jun-Hyung
AU - Kim, Ji-In
AU - Kwon, Goo-Rak
TI - APOE ε4 as a predictor of cognitive decline and its interaction with hippocampal volume in Alzheimer's disease
T2 - Frontiers in aging neuroscience
J2 - Front Aging Neurosci
PY - 2026
DA - 2026/04/22
VL - 18
SP - 1730265
SN - 1663-4365
PB - Frontiers Media SA
DO - 10.3389/fnagi.2026.1730265
UR - https://doi.org/10.3389/fnagi.2026.1730265
LA - en
ER -

CSL-JSON

{
"id": "10.3389/fnagi.2026.1730265",
"type": "article-journal",
"title": "APOE ε4 as a predictor of cognitive decline and its interaction with hippocampal volume in Alzheimer's disease",
"container-title": "Frontiers in aging neuroscience",
"author": [
{
"family": "Khan",
"given": "Faizaan Fazal"
},
{
"family": "Kim",
"given": "Jun-Hyung"
},
{
"family": "Kim",
"given": "Ji-In"
},
{
"family": "Kwon",
"given": "Goo-Rak"
}
],
"container-title-short": "Front Aging Neurosci",
"volume": "18",
"page": "1730265",
"DOI": "10.3389/fnagi.2026.1730265",
"PMID": "42100481",
"PMCID": "PMC13144130",
"ISSN": "1663-4365",
"publisher": "Frontiers Media SA",
"URL": "https://doi.org/10.3389/fnagi.2026.1730265",
"language": "en",
"issued": {
"date-parts": [
[
2026,
4,
22
]
]
}
}

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

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1038/s41398-026-04143-x [code]
Neighborhood opportunity, white matter, and cognition in a large pediatric neuroimaging study.
Journal: Translational psychiatry
In common: statsmodels, seaborn, pandas, 3 other tools, structural MRI / diffusion, 1 reference
[2] doi:10.1016/j.cell.2026.05.047 [code]
An emergent disease-associated motor neuron state precedes cell death in ALS.
Journal: Cell
In common: statsmodels, seaborn, pandas, 3 other tools, genetics / omics, 1 reference
[3] doi:10.1038/s41586-026-10629-x [code]
Whole-genome duplication shaped cell-type evolution in the vertebrate brain.
Journal: Nature
In common: statsmodels, seaborn, pandas, 3 other tools, genetics / omics, 1 reference
[4] doi:10.1016/j.isci.2026.116825 [code]
Social hierarchy shapes behavioral and transcriptional responses to chronic stress and ketamine in male mice.
Journal: iScience
In common: statsmodels, seaborn, pandas, 3 other tools, genetics / omics, 1 reference
[5] doi:10.1038/s41514-026-00390-w [code]
Mild cognitive impairment cases affect the predictive power of Alzheimer's disease diagnostic models using routine clinical variables.
Journal: npj aging
In common: seaborn, pandas, SciPy, 2 other tools, Alzheimer's / dementia, clinical / translational, 1 reference
[6] doi:10.1073/pnas.2516601123 [code]
Unveiling the glymphatic system's role in brain aging: A comprehensive biomarker and modifiable intervention target.
Journal: Proceedings of the National Academy of Sciences of the United States of America
In common: seaborn, pandas, SciPy, 2 other tools, structural MRI / diffusion, clinical / translational, 1 reference
[7] doi:10.1186/s40708-026-00316-y [code]
Generalizable and explainable deep learning for brain MRI: a multi-cohort evaluation of 3D architectures for age and sex prediction.
Journal: Brain informatics
In common: seaborn, pandas, SciPy, 2 other tools, structural MRI / diffusion, 1 reference
[8] doi:10.1162/imag.a.1278 [code]
Young and old adult brains experience opposite effects of acute sleep restriction on the functional connectivity network.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: statsmodels, seaborn, pandas, 3 other tools, 1 reference
[9] doi:10.1093/cercor/bhag077 [code]
The longitudinal development of intrinsic timescales in infancy and their relation to alpha brain rhythm.
Journal: Cerebral cortex (New York, N.Y. : 1991)
In common: statsmodels, seaborn, pandas, 3 other tools, 1 reference
[10] doi:10.1162/imag.a.1246 [code]
A 3.5-minute-long reading-based fMRI localizer for the language network.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: statsmodels, seaborn, pandas, 3 other tools, 1 reference

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.