OSCR

LPS-induced inflammation differentially affects endogenous Ca<sup>2</sup>⁺ activity in mouse and human iPSC-derived astrocytes.

Code ↔ Paper

1 match 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 1 match
  1. [1] § Methods › LDA, component analysis and weight matrix generation ↔ 2026-06-05_LPS_Paper_Figs_GS_AZ_KL.ipynb, lines 119–263 · score 0.57 · MaxSize, Negative Matrix Factorization, weights, distance, components, NMF

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 · 640 lines · 23 KB · GPL-3.0 · 1 match

  1. # %% [markdown]
  2. # ## LPS-induced inflammation differentially affects endogenous Ca2⁺ activity in mouse and human iPSC-derived astrocytes
  3. # ### Franziska E. Müller, Flavian Ivanov, Anne-Catharine Studt, Ida Nitzsche, Frauke S. Bahr, Anna-Lena Krüger, Josephine Labus, Ghanendra Singh, Evgeni G. Ponimaskin, Kerstin Lenk* & Andre Zeug*
  4. # \* authors contributed equally
  5. # Mol Med 32, 52 (2026). https://doi.org/10.1186/s10020-026-01450-3
  6. #
  7. # **Contribution to code:** Ghanendra Singh, Andre Zeug, Kerstin Lenk
  9. # %% [markdown]
  10. # **Install libraries if needed:**
  11. # pip install seaborn, used version seaborn-0.13.2
  12. # pip install scikit-learn, used version scikit-learn-1.9.0
  13. # pip install openpyxl, used version openpyxl-3.1.5
  15. # %%
  16. # Import Library
  17. import pandas as pd
  18. import numpy as np
  19. import matplotlib.pyplot as plt
  20. from matplotlib import cm, ticker
  21. import matplotlib.colors as colors
  22. %matplotlib inline
  23. import seaborn as sns
  24. import os
  26. from matplotlib.backends.backend_pdf import PdfPages
  27. from datetime import datetime
  28. from matplotlib.colors import Normalize
  30. # ML Methods
  31. from sklearn.manifold import TSNE
  32. from sklearn.decomposition import PCA, FastICA, NMF
  33. # Importing different scalars
  34. from sklearn.preprocessing import StandardScaler,MinMaxScaler
  36. # Handy to ignore warnings
  37. import warnings
  38. warnings.filterwarnings("ignore")
  40. # %% [markdown]
  41. # ### Load data from Excel file
  42. # #### Note:
  43. # * Place your excel file in the same folder as jupyter notebook or use custom path
  44. # * Ignores the first header row for simplicity
  45. # * Index starts from zero
  47. # %%
  48. # Load data from the excel file into a dataframe using pandas pd
  49. path = 'MLData_comp_LPS_Treatment_Ch2_updated_file.xlsx'
  50. xl = pd.ExcelFile(path)
  51. # Check multipe sheets in the excel file #xl.sheet_names
  52. df_b18_g_5 = xl.parse(1,header=1) # pass sheet index as 1 for 18bin with gamma 0.5 excel data
  54. # %%
  55. # Visualize first five observations
  56. df_b18_g_5.head()
  58. # %% [markdown]
  59. # ### Dataframe with multi Index
  60. # * Data columm contains (first five sub columns)
  61. # * Here it is fixed (metadata size = 5)
  63. # %%
  64. # Function to load data with proper columns names
  65. # Default binsize in the function is 18
  67. def load_data(df,bs=18): # bs is bin size in short
  68. md = 5 #md = 5 as metadata size
  69. df_xl = pd.concat([df.iloc[:,0:md],
  70. df.iloc[:,md:1*bs*6+md],
  71. df.iloc[:,1*bs*6+md:2*bs*6+md],
  72. df.iloc[:,2*bs*6+md:3*bs*6+md],
  73. df.iloc[:,3*bs*6+md:4*bs*6+md]],
  74. axis=1,
  75. keys=['Data','maxArea','duration','maxDist','slopeCa'])
  76. return df_xl
  78. # %% [markdown]
  79. # ### Load the whole data file (HIP, PFC + iPSCs)
  80. # * Add the path to iPSCs data
  82. # %%
  83. path2 = 'MLData_hiPSC-Astros-AAV-V5-basal_Ch2.xlsx'
  85. xl2 = pd.ExcelFile(path2)
  86. df_b18_g_5_iPSC = xl2.parse(0,header=1)
  87. df = load_data(df_b18_g_5_iPSC, binsize)
  88. df.head()
  90. # %% [markdown]
  91. # #### Data representation can also be further simplified as
  92. # * Metadata (First five columns) or more for later datasets
  93. # * Main data (More no. of features can be considered)
  94. # * Generally a single name is used to just modify binsize and run whole cells at once.
  95. # * Dataframes are named with bin sizes to ensure to choose right binsize and dataframe.
  97. # %% [markdown]
  98. # ### Main dataframe
  100. # %%
  101. # Always use this bin size
  102. binsize = 18 #default binsize
  103. # change dataframe and binsize for different data
  104. df = load_data(df_b18_g_5, binsize)
  106. # %% [markdown]
  107. # ### Plotting different x-axis for bin size 18 for flame plots
  109. # %%
  110. bins = 18 # binsize
  111. bins_area = np.logspace(np.log10(0.44), np.log10(4000), num=bins, base=10)
  112. bins_duration = np.logspace(np.log10(0.6), np.log10(300), num=bins, base=10)
  113. bins_distance = np.logspace(np.log10(0.5), np.log10(100), num=bins, base=10)
  114. bins_slope = np.logspace(np.log10(0.001), np.log10(8), num=bins, base=10)
  116. x_axis_bins = [bins_area,bins_duration,bins_distance,bins_slope]
  117. y_axis_thresholds = [.5,1,2,5,10,20]
  119. # %% [markdown]
  120. # ## NMF (Non-Negative Matrix Factorization)
  122. # %% [markdown]
  123. # ### On entire dataset
  125. # %%
  126. # Load data
  127. from matplotlib.backends.backend_pdf import PdfPages
  128. from datetime import datetime
  130. def nmf_whole(nmf_Comp=6, plot_contour_flag=True, region="Whole_data"):
  132. # df = load_data(df_b18_g_5)
  133. df = load_data(df_b18_g_5, 18)
  134. features = ["maxArea", "duration", "maxDist", "slopeCa"]
  135. legend_names = ["Ctrl.", "3 h LPS", "24 h LPS"]
  136. if region == "HIP":
  137. dfc = df[df["Data"]["Condition"] <= 3]
  139. elif region == "PFC":
  140. dfc = df[df["Data"]["Condition"] > 3]
  141. elif region == "iPSC":
  142. dfc = load_data(df_b18_g_5_iPSC, 18)
  143. print("iPSC data loaded")
  144. else:
  145. dfc = df
  147. X = dfc.iloc[:, 5:] # data (ignore first five columns)
  148. y = dfc.iloc[:, 0] # labels
  149. nmfC = nmf_Comp # of NMF components
  150. nF = 4 # no. of features
  152. nmf = NMF(n_components=nmfC, init="random", random_state=0, tol=1e-3)
  153. W = nmf.fit_transform(X)
  154. H = nmf.components_
  155. # Normalize H to have a maximum value of 1 and adjust W accordingly
  156. W = W * H.max(axis=1)
  157. H = H / H.max(axis=1, keepdims=True)
  158. # Manually reorganize matrices to fit color sceme
  159. if region == "HIP":
  160. if nmf_Comp == 2:
  161. H = H[[1, 0], :]
  162. W = W[:, [1, 0]]
  163. # elif nmf_Comp == 3:
  164. # H = H[[0, 2, 1], :]
  165. # W = W[:, [0, 2, 1]]
  166. elif region == "iPSC":
  167. # if nmf_Comp == 2:
  168. # H = H[[1, 0], :]
  169. # W = W[:, [1, 0]]
  170. if nmf_Comp == 3:
  171. H = H[[0, 2, 1], :]
  172. W = W[:, [0, 2, 1]]
  173. # print(H[:5, :5])
  174. # print(H_reorganized[:5, :5])
  177. H = H.reshape(nmfC, 4, 6, 18)
  178. vmin = 0
  179. vmax = 1 # H.max()
  180. levels = np.linspace(vmin, vmax, 11) # Define levels for contourf
  181. levelsC = np.linspace(0.2, vmax, 6) # Define levels for contour
  182. norm = colors.Normalize(vmin=vmin, vmax=vmax)
  184. # Generate a timestamp
  185. timestamp = datetime.now().strftime("%Y-%m-%d_%H.%M.%S")
  187. # Create a PdfPages object with a timestamped file name
  188. file_name = f'tmp/NMF_{region}_{nmf_Comp}_{timestamp}.pdf'
  189. file_name_W = f'tmp/NMF_{region}_{nmf_Comp}_{timestamp}_W.csv'
  190. # Convert W to a DataFrame
  191. W_df = pd.DataFrame(W)
  192. # Save the DataFrame to a CSV file
  193. W_df.to_csv(file_name_W, index=False)
  194. #print("W matrix saved as:", file_name_W)
  195. # Set the default font size
  196. plt.rcParams.update({'font.size': 16}) # Change 16 to your desired font size
  198. feature_labels = [r"MaxSize [$\mu$m$^2$]",
  199. "Duration [s]",
  200. r"Distance [$\mu$m]",
  201. r"Ca$^{2+}$ change [a.u.]"
  202. ]
  204. legend_names = ["Ctrl.", "3 h LPS", "24 h LPS"]
  206. with PdfPages(file_name) as pdf:
  208. ## Contours overlayed on top of each other
  209. cmap = ["Greys", "Greens", "Reds", "Blues", "Purples", "Oranges"]
  210. res = plt.subplots(1, nF+1, figsize=(18, 4),facecolor='none')
  211. axes : list[plt.Axes] = res[1] # Initialize axes list
  212. fig2 = res[0]
  213. for f in range(nF):
  214. sax = plt.subplot(1, nF+1, f + 1)
  215. for nmf_i in range(nmfC):
  216. contour = plt.contour(x_axis_bins[f], range(0, 6, 1), H[nmf_i, f], levels=levelsC,
  217. vmin=vmin, vmax=vmax, extend="neither", cmap=cmap[nmf_i],
  218. alpha=0.7, linewidths=3, norm=norm)
  219. # fp.cmap.set_over('black')
  220. # fp.cmap.set_under('white')
  221. plt.xscale("log") # Set x-axis to logarithmic scale
  222. plt.yticks(range(0, 6, 1), y_axis_thresholds)
  223. plt.xlim(x_axis_bins[f].min(), x_axis_bins[f].max()) # Set x-axis limits
  224. # plt.autoscale(enable=True, axis='x', tight=True)
  225. # contour.set_clim(vmin, vmax) # Set colorbar limits explicitly
  226. contour.cmap.set_under("white")
  227. # plt.title(features[f], pad=20)
  228. plt.ylabel("Threshold (NMF Comp. %d)" % (nmf_i + 1),fontsize=15)
  229. plt.xlabel(feature_labels[f],fontsize=15)
  230. sax.patch.set_alpha(0) # Make the axes background transparent
  231. plt.tight_layout()
  232. sax.set_aspect(aspect="auto") # Options: 'equal', 'auto', or a numeric value
  233. # plt.show()
  235. # Weight distribution
  236. cmap=["grey","green","red","blue","purple","orange"]
  237. clabel = dfc["Data"]["Cond.Label"].unique()
  238. cond_u = dfc["Data"]["Condition"].unique()
  239. cond_u_int = list(map(int, cond_u))
  240. WW = np.zeros((cond_u.shape[0], nmfC))
  241. list_cond = []
  242. for i,cond_i in enumerate(cond_u):
  243. W_i = W[dfc["Data"]["Condition"] == cond_i, :]
  244. list_cond.append(W_i)
  245. W_i = np.mean(W_i, axis=0)
  246. WW[i, :] = W_i
  247. # list_cond[0][:,0]
  248. # plt.figure(figsize=(12, 6))
  250. plt.subplot(1, nF+1, nF + 1)
  251. for i in range(WW.shape[1]): # Loop over the number of NMF components
  252. plt.plot(cond_u,WW[:,i],ls="--",marker="s",
  253. label=f"NMF Comp{i+1}", color=cmap[i % len(cmap)]) # Use colors from cmap
  254. # plt.legend()
  255. plt.title("Weight Distribution", pad=20)
  256. plt.xticks(cond_u,legend_names)
  257. plt.xlabel("Condition Index")
  258. plt.ylabel("Weight")
  259. plt.ylim(bottom=0)
  260. plt.gca().patch.set_alpha(0) # Make the axes background transparent#
  261. plt.savefig("tmp/NMF_Weight_Distribution_%d.svg"%nmfC, dpi=300, bbox_inches='tight', transparent=True)
  262. plt.show()
  263. #pdf.savefig(fig2, transparent=True) # Save the current figure to the PDF
  266. # %%
  267. from datetime import datetime
  269. def nmf_whole(nmf_Comp=6, plot_contour_flag=True, region="Whole_data",suptext="", suptextcolor=""):
  271. # Load data
  272. df = load_data(df_b18_g_5, 18)
  274. if region == "HIP":
  275. dfc = df[df["Data"]["Condition"] <= 3]
  277. elif region == "PFC":
  278. dfc = df[df["Data"]["Condition"] > 3]
  280. elif region == "iPSC":
  281. dfc = load_data(df_b18_g_5_iPSC, 18)
  282. #print("iPSC data loaded")
  284. else:
  285. dfc = df
  287. # Labels
  288. feature_labels = [r"MaxSize [$\mu$m$^2$]","Duration [s]",r"Distance [$\mu$m]",r"Ca$^{2+}$ change [a.u.]",]
  289. legend_names = ["Ctrl.", "3 h LPS", "24 h LPS"]
  291. # NMF
  292. X = dfc.iloc[:, 5:]
  293. nmf = NMF(n_components=nmf_Comp,init="random",random_state=0,tol=1e-3,)
  295. W = nmf.fit_transform(X)
  296. H = nmf.components_
  298. # Normalize
  299. W = W * H.max(axis=1)
  300. H = H / H.max(axis=1, keepdims=True)
  302. # Optional component reordering
  303. if region == "HIP" and nmf_Comp == 2:
  304. H = H[[1, 0], :]
  305. W = W[:, [1, 0]]
  307. elif region == "iPSC" and nmf_Comp == 3:
  308. H = H[[0, 2, 1], :]
  309. W = W[:, [0, 2, 1]]
  311. # Save W matrix
  312. timestamp = datetime.now().strftime("%Y-%m-%d_%H.%M.%S")
  313. pd.DataFrame(W).to_csv(f"tmp/NMF_{region}_{nmf_Comp}_{timestamp}_W.csv",index=False,)
  315. # Reshape H
  316. H = H.reshape(nmf_Comp, 4, 6, 18)
  318. vmin, vmax = 0, 1
  319. levelsC = np.linspace(0.2, vmax, 6)
  321. norm = colors.Normalize(vmin=vmin, vmax=vmax)
  323. # Figure layout
  324. plt.rcParams.update({"font.size": 14})
  326. fig, axes = plt.subplots(1,5,figsize=(16, 4),constrained_layout=True,)
  328. contour_cmaps = ["Greys","Greens","Reds","Blues","Purples","Oranges",]
  330. # Contour plots (first 4 columns)
  331. y_axis_thresholds = [0.5, 1, 2, 5, 10, 20]
  332. for f in range(4):
  334. ax = axes[f]
  335. for nmf_i in range(nmf_Comp):
  337. contour = ax.contour(x_axis_bins[f],range(6),H[nmf_i, f],levels=levelsC,cmap=contour_cmaps[nmf_i],
  338. alpha=0.7,linewidths=3,norm=norm,)
  340. contour.cmap.set_under("white")
  342. ax.set_xscale("log")
  343. ax.set_xlim(x_axis_bins[f].min(),x_axis_bins[f].max(),)
  344. ax.set_xlabel(feature_labels[f])
  346. ax.set_yticks(range(6))
  348. if f == 0:
  349. ax.set_yticks(range(6))
  350. ax.set_yticklabels(y_axis_thresholds)
  351. ax.tick_params(axis="y", labelleft=True)
  352. ax.set_ylabel(r"$Ca^{2+}$ threshold")
  355. # Force x ticks to show on log scale
  356. ax.tick_params(axis="x", which="both", labelbottom=True)
  357. ax.set_aspect("auto")
  359. fig.supylabel(suptext, color=suptextcolor,fontsize=16, fontweight="bold")
  361. # Weight distribution (last column)
  362. ax_w = axes[4]
  364. cmap_weights = ["grey","green","red","blue","purple","orange",]
  365. cond_u = dfc["Data"]["Condition"].unique()
  367. WW = np.zeros((len(cond_u), nmf_Comp))
  368. for i, cond_i in enumerate(cond_u):
  369. W_i = W[dfc["Data"]["Condition"] == cond_i]
  370. WW[i] = np.mean(W_i, axis=0)
  372. for i in range(nmf_Comp):
  374. ax_w.plot(cond_u,WW[:, i],ls="--",marker="s",color=cmap_weights[i % len(cmap_weights)],label=f"NMF {i + 1}",)
  376. ax_w.set_xticks(cond_u)
  377. ax_w.set_xticklabels(legend_names)
  379. ax_w.set_xlabel("Condition")
  380. ax_w.set_ylabel("Weight")
  381. ax_w.set_ylim(bottom=0)
  382. ax_w.tick_params(axis="x", which="both", labelbottom=True)
  383. ax_w.tick_params(axis="y", which="both", labelleft=True)
  385. ax_w.legend(frameon=False)
  387. # Export
  388. #plt.savefig(f"tmp/NMF_{region}_{nmf_Comp}.svg",dpi=300,bbox_inches="tight",transparent=True)
  390. plt.show()
  392. # %%
  393. os.makedirs("tmp", exist_ok=True)
  394. nmf_whole(nmf_Comp=3, plot_contour_flag=True, region="HIP", suptext="HC", suptextcolor="#2F4F8F")
  395. nmf_whole(nmf_Comp=3, plot_contour_flag=True, region="PFC", suptext="PFC",suptextcolor="#A01B8E")
  396. nmf_whole(nmf_Comp=3, plot_contour_flag=True, region="iPSC", suptext="PFC",suptextcolor="#18C7A0")
  398. # %% [markdown]
  399. # ## PCA, NMF and LDA
  401. # %%
  402. from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
  403. def lda_compare(region="Whole_data"):
  405. # df = load_data(df_b18_g_5)
  406. features = ["maxArea", "duration", "maxDist", "slopeCa"]
  407. legend_names = ["Ctrl.", "3 h LPS", "24 h LPS"]
  409. if region == "HIP":
  410. df = load_data(df_b18_g_5, 18)
  411. dfc = df[df["Data"]["Condition"] <= 3]
  412. colors_rgb = [(144, 191, 249), (139, 144, 206), (0, 0, 192)] # RGB colors
  413. elif region == "PFC":
  414. df = load_data(df_b18_g_5, 18)
  415. dfc = df[df["Data"]["Condition"] > 3]
  416. colors_rgb = [(255, 160, 160), (193, 139, 183), (128, 0, 128)] # RGB colors
  417. elif region == "iPSC":
  418. dfc = load_data(df_b18_g_5_iPSC, 18)
  419. print("iPSC data loaded")
  420. colors_rgb = [(169, 254, 169), (15, 153, 178), (3, 78, 97)] # RGB colors
  421. else:
  422. df = load_data(df_b18_g_5, 18)
  423. dfc = df
  425. colors_rgb = np.array(colors_rgb) / 255.0 # Normalize RGB values to [0, 1]
  426. X = dfc.iloc[:, 5:] # data (ignore first five columns)
  427. y = dfc.iloc[:, 0] # labels
  429. noC = 2 # no. of components
  431. lda = LDA(n_components=noC) #,shrinkage='Auto',solver='lsqr',priors=None,store_covariance=False, tol=0.0001
  432. pca = PCA(n_components=noC,random_state=22)
  434. #nmf = NMF(n_components=noC)
  435. nmf = NMF(n_components=noC, init='random', random_state=22)
  437. Xs = StandardScaler().fit_transform(X)
  438. Xs = MinMaxScaler().fit_transform(X)
  440. X_lda = lda.fit_transform(X,y)
  441. X_pca = pca.fit_transform(X)
  442. X_nmf = nmf.fit_transform(X)
  444. # Define RGB colors for each condition
  445. unique_conditions = np.unique(y)
  446. color_map = {condition: colors_rgb[i % len(colors_rgb)] for i, condition in enumerate(unique_conditions)}
  447. condition_colors = [color_map[condition] for condition in y]
  449. # Generate a timestamp
  450. timestamp = datetime.now().strftime("%Y-%m-%d_%H.%M.%S")
  451. # Create a PdfPages object with a timestamped file name
  452. file_name = f'tmp/LDA_{region}.pdf' #_{timestamp}
  453. with PdfPages(file_name) as pdf:
  454. res = plt.subplots(1, 3, figsize=(16, 4), constrained_layout=False)
  455. axes : list[plt.Axes] = res[1] # Initialize axes list
  456. fig = res[0]
  457. print("Default font size:", plt.rcParams['font.size'])
  458. adjusted_font_size = plt.rcParams['font.size']-2 # Adjust font size for legends
  460. # PCA scatter plot
  461. for i, condition in enumerate(unique_conditions):
  462. mask = y == condition
  463. axes[0].scatter(X_pca[mask, 0], X_pca[mask, 1], color=color_map[condition], label=legend_names[i])
  464. axes[0].set_title('PCA')
  465. axes[0].set_xlabel('PCA 1')
  466. axes[0].set_ylabel('PCA 2')
  467. axes[0].set_aspect('equal') # Set equal scaling
  468. axes[0].legend(fontsize=adjusted_font_size)
  470. # NMF scatter plot
  471. for i, condition in enumerate(unique_conditions):
  472. mask = y == condition
  473. axes[1].scatter(X_nmf[mask, 0], X_nmf[mask, 1], color=color_map[condition], label=legend_names[i])
  474. axes[1].set_title('NMF')
  475. axes[1].set_xlabel('NMF 1')
  476. axes[1].set_ylabel('NMF 2')
  477. axes[1].set_aspect('equal') # Set equal scaling
  478. axes[1].legend(fontsize=adjusted_font_size)
  480. # LDA scatter plot
  481. for i, condition in enumerate(unique_conditions):
  482. mask = y == condition
  483. axes[2].scatter(X_lda[mask, 0], X_lda[mask, 1], color=color_map[condition], label=legend_names[i])
  484. axes[2].set_title('LDA')
  485. axes[2].set_xlabel('LDA 1')
  486. axes[2].set_ylabel('LDA 2')
  487. axes[2].set_aspect('equal') # Set equal scaling , adjustable='box'
  488. axes[2].legend(fontsize=adjusted_font_size)
  490. # Add legend
  491. axes[2].legend(loc='best', fontsize='small', framealpha=0.5, borderaxespad=0.2,handletextpad=0.2) #, edgecolor='black'
  492. #axes[2].legend(loc='center left', bbox_to_anchor=(1.05, 0.5), fontsize='small', framealpha=0.5, borderaxespad=0.2) #, edgecolor='black'
  494. # Adjust layout for better spacing
  495. plt.tight_layout()
  496. plt.show()
  497. pdf.savefig(fig)
  499. # %%
  500. lda_compare('HIP')
  502. # %%
  503. lda_compare('PFC')
  505. # %% [markdown]
  506. # ## Supplementary Figure 8
  508. # %%
  509. # load data
  510. from matplotlib.backends.backend_pdf import PdfPages
  511. from datetime import datetime
  512. from matplotlib.colors import Normalize
  514. def nmf_whole_modified(nmf_Comp=6, plot_contour_flag=True, region="Whole_data"):
  516. # df = load_data(df_b18_g_5)
  517. df = load_data(df_b18_g_5, 18)
  518. features = ["maxArea", "duration", "maxDist", "slopeCa"]
  519. legend_names = ["Ctrl.", "3 h LPS", "24 h LPS"]
  520. if region == "HIP":
  521. dfc = df[df["Data"]["Condition"] <= 3]
  523. elif region == "PFC":
  524. dfc = df[df["Data"]["Condition"] > 3]
  525. elif region == "iPSC":
  526. dfc = load_data(df_b18_g_5_iPSC, 18)
  527. print("iPSC data loaded")
  528. else:
  529. dfc = df
  531. X = dfc.iloc[:, 5:] # data (ignore first five columns)
  532. y = dfc.iloc[:, 0] # labels
  533. nmfC = nmf_Comp # of NMF components
  534. nF = 4 # no. of features
  536. nmf = NMF(n_components=nmfC, init="random", random_state=0, tol=1e-3)
  537. W = nmf.fit_transform(X)
  538. H = nmf.components_
  539. # Normalize H to have a maximum value of 1 and adjust W accordingly
  540. W = W * H.max(axis=1)
  541. H = H / H.max(axis=1, keepdims=True)
  542. # Manually reorganize matrices to fit color sceme
  543. if region == "HIP":
  544. if nmf_Comp == 2:
  545. H = H[[1, 0], :]
  546. W = W[:, [1, 0]]
  547. # elif nmf_Comp == 3:
  548. # H = H[[0, 2, 1], :]
  549. # W = W[:, [0, 2, 1]]
  550. elif region == "iPSC":
  551. # if nmf_Comp == 2:
  552. # H = H[[1, 0], :]
  553. # W = W[:, [1, 0]]
  554. if nmf_Comp == 3:
  555. H = H[[0, 2, 1], :]
  556. W = W[:, [0, 2, 1]]
  557. # print(H[:5, :5])
  558. # print(H_reorganized[:5, :5])
  561. H = H.reshape(nmfC, 4, 6, 18)
  562. vmin = 0
  563. vmax = 1 # H.max()
  564. levels = np.linspace(vmin, vmax, 11) # Define levels for contourf
  565. levelsC = np.linspace(0.2, vmax, 6) # Define levels for contour
  566. norm = colors.Normalize(vmin=vmin, vmax=vmax)
  568. # Generate a timestamp
  569. timestamp = datetime.now().strftime("%Y-%m-%d_%H.%M.%S")
  570. # Create a PdfPages object with a timestamped file name
  571. file_name = f'tmp/NMF_{region}_{nmf_Comp}_{timestamp}.pdf'
  572. file_name_W = f'tmp/NMF_{region}_{nmf_Comp}_{timestamp}_W.csv'
  573. # Convert W to a DataFrame
  574. W_df = pd.DataFrame(W)
  576. # Save the DataFrame to a CSV file
  577. #W_df.to_csv(file_name_W, index=False)
  578. #print("W matrix saved as:", file_name_W)
  580. # Set the default font size
  581. plt.rcParams.update({'font.size': 12}) # Change 16 to your desired font size
  582. plt.rcParams['xtick.labelsize']=12
  583. plt.rcParams['xtick.bottom'] = True
  584. plt.rcParams['xtick.labelbottom'] = False
  585. plt.rcParams['ytick.left'] = True
  586. plt.rcParams['ytick.labelleft'] = False
  587. plt.rcParams['ytick.labelsize']=12
  589. with PdfPages(file_name) as pdf:
  590. #fig1 = plt.figure(figsize=(18,12))
  591. res = plt.subplots(nmfC,nF, figsize=(18,nmfC*4),sharex=False, sharey=True)
  592. axes1 : list[plt.Axes] = res[1] # Initialize axes list
  593. fig1 = res[0]
  595. plt.suptitle("# NMF components: %d"%nmfC, y=0.92) # Adjust the y parameter for more space
  597. #plt.subplots_adjust(top=0.95) # Adjust the top spacing for more gap
  598. for nmf_i in range(nmfC):
  599. for f in range(nF):
  600. plt.subplot(nmfC,nF,nF*nmf_i +f+1)
  601. if plot_contour_flag:
  602. contour = plt.contourf(x_axis_bins[f],range(0,6,1),H[nmf_i,f],cmap='jet',levels=levels,
  603. vmin=vmin,vmax=vmax, extend='neither')
  604. else:
  605. contour = plt.imshow(H[nmf_i,f], interpolation='bicubic',aspect='auto',
  606. origin='lower',cmap='jet',vmin=0)
  608. contour.set_clim(vmin, vmax) # Set colorbar limits explicitly
  610. if f==0:
  611. plt.tick_params(left=True,labelleft=True)
  612. plt.ylabel(r'$Ca^{2+}$ Threshold')
  614. if nmf_i == nmfC-1:
  615. plt.tick_params(bottom=True,labelbottom=True)
  616. plt.xlabel(features[f])
  618. #fp.cmap.set_over('black')
  619. #fp.cmap.set_under('white')
  621. plt.xscale('log') # Set x-axis to logarithmic scale
  622. plt.yticks(range(0,6,1),y_axis_thresholds,fontsize=15)
  623. #plt.xlim(x_axis_bins[f].min(), x_axis_bins[f].max()) # Set x-axis limits
  624. plt.xticks(fontsize=12)
  625. plt.yticks(fontsize=12)
  626. plt.autoscale(enable=True, axis='x', tight=True)
  628. plt.subplots_adjust(hspace=0.04,wspace=0.04)
  629. cbar = fig1.colorbar(contour,ax=axes1.ravel().tolist(),shrink=0.5,pad=0.02) # ,fontsize=15
  630. cbar.mappable.set_clim(vmin, vmax) # Set colorbar limits explicitly
  631. #fig1.tight_layout()
  633. plt.gca().patch.set_alpha(0) # Make the axes background transparent#
  634. pdf.savefig(fig1) # Save the current figure to the PDF
  635. #plt.savefig("NMF_Components_%d.pdf"%nmfC, dpi=300, bbox_inches='tight')
  636. plt.savefig('NMF_Components_%d_iPSC.svg'%nmfC,format="svg", dpi=600, bbox_inches='tight',transparent=True)
  637. plt.show()
  639. n_nmfc=3
  640. nmf_whole_modified(n_nmfc,plot_contour_flag=True,region = "HIP")

2026-06-05_LPS_Paper_Figs_GS_AZ_KL.ipynb at commit 1e9979f, under GPL-3.0 · at the source

Overview

Authors: Franziska E Müller1, Flavian Ivanov1, Anne-Catharine Studt1, Ida Nitzsche1, Frauke S Bahr1, Anna-Lena Krüger1, Josephine Labus1, Ghanendra Singh2, Evgeni G Ponimaskin1, Kerstin Lenk2, Andre Zeug1
ORCID iDs: Franziska E Müller, Frauke S Bahr, Josephine Labus, Kerstin Lenk, Andre Zeug
  1. Cellular Neurophysiology, Institute of Neurophysiology, Hannover Medical School, Hannover, Germany
  2. Institute of Neural Engineering, Graz University of Technology, Graz, Austria
Institutions: Medizinische Hochschule Hannover (Germany); Graz University of Technology (Austria)
Journal: Molecular medicine (Cambridge, Mass.), volume 32, issue 1, article 52
Dates: received 15 August 2025; accepted 26 February 2026; published online 5 March 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1186/s10020-026-01450-3 · PMID 41787269 · PMCID PMC13064076 · OpenAlex W7133828055
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: human (organism), mouse (organism), cellular / molecular (subfield)
Methods: Statistics, Machine learning, fMRI & imaging
Keywords: Astrocytes, Lipopolysaccharide, Ca2+ signalling, Calcium, Heterogeneity
MeSH: Astrocytes*, Calcium*, Induced Pluripotent Stem Cells*, Inflammation*, Lipopolysaccharides*, Animals, Calcium Signaling, Cells, Cultured, Humans, Mice (* major topic)
Topic: Protein Kinase Regulation and GTPase Signaling (Molecular Biology, Biochemistry, Genetics and Molecular Biology), according to OpenAlex
Funding: Medizinische Hochschule Hannover (MHH)
Citations: cited by 1 paper (Europe PMC); 58 references in the paper

Abstract

Background: Mouse and human astrocytes exhibit substantial species-specific differences in both morphology and function. Their response to inflammatory stimuli, however, remains underexplored despite being crucial for understanding bidirectional astrocyte-neuron signaling dynamics and for translating preclinical findings to human-relevant applications. Induced pluripotent stem cell-based models thus offer a powerful platform to investigate these mechanisms in the context of the human neural connectome.

Methods: We apply two well-established in vitro protocols by exposing cultured astrocytes to lipopolysaccharide (LPS) for either 3 or 24 h to trigger an inflammatory response. We investigated how LPS-induced inflammation affects the endogenous Ca2+ activity in astrocytes derived from the mouse hippocampus (HC) and prefrontal cortex (PFC), as well as human induced pluripotent stem cell (hiPSC)-derived astrocytes. Both, morphological changes and Ca2+ activity were analyzed using the volume fraction (VF) approach and our previously developed multi-threshold event detection (MTED) combined with machine learning-driven non-negative matrix factorization (NMF).

Results: The comprehensive assessment of Ca2+ activity patterns and their relation to cell morphology revealed significant alterations in response to LPS treatment, and further between mouse and human hiPSC-derived astrocytes. While both mouse and human astrocytes show increased Ca2+ event frequency after short-term LPS exposure, after 24 h of LPS treatment Ca2+ activity is severely restricted in PFC astrocytes but substantially increased in human astrocytes.

Conclusions: Our findings highlight the unique properties of human iPSC-derived astrocytes and provide detailed insights into how Ca2+ signaling becomes dysregulated under neuroinflammatory conditions. Understanding the species-specific responses is essential for advancing stem cell-based models of human astrocyte-neuron signaling circuits and for developing targeted therapeutic strategies to alleviate neuroinflammation and Ca2+-related dysregulation in neurological diseases.

Supplementary Information: The online version contains supplementary material available at 10.1186/s10020-026-01450-3.

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 1 match between paragraphs and lines of code.

kerstinlenk/QuantifyCalciumDynamics

License: GPL-3.0
State: the link answers, verified on 30 September 2026
Evidence: files inventoried
Commit: 1e9979fdbc30fc6f47736777ae562e15338a2aa6, 7 June 2026
Languages: Jupyter (1)
Size: 3 files, 1 script
Software Heritage: not archived
Found in: “Data availability”
Holds: README, license file, 1 notebook
Not found: CITATION.cff, environment file, tests, continuous integration, documentation
Tools: Matplotlib (1 file), NumPy (1 file), pandas (1 file), scikit-learn (1 file), seaborn (1 file)
Availability: 1 check, the latest on 30 September 2026: the link answers
  • 30 September 2026: the link answers
3 files
At the source: github.com/kerstinlenk/QuantifyCalciumDynamics

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;
  • 1 script, each with its path and the digest of its content;
  • 1 match 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

Datasets cited

Data availability

The data and MATLAB code developed in this study are available from the corresponding author upon request. The Python code is available on GitHub using the following link: https://github.com/kerstinlenk/QuantifyCalciumDynamics.

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, 30 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 11 authors, 5 keywords, 10 MeSH terms, 1 funder, 57 references.

Cite

This paper

Müller, F. E., Ivanov, F., Studt, A.-C., Nitzsche, I., Bahr, F. S., Krüger, A.-L., Labus, J., Singh, G., Ponimaskin, E. G., Lenk, K., & Zeug, A. (2026). LPS-induced inflammation differentially affects endogenous Ca&lt;sup&gt;2&lt;/sup&gt;⁺ activity in mouse and human iPSC-derived astrocytes. Molecular medicine (Cambridge, Mass.), 32(1), 52. https://doi.org/10.1186/s10020-026-01450-3

BibTeX

@article{muller2026lps,
author = {Müller, Franziska E and Ivanov, Flavian and Studt, Anne-Catharine and Nitzsche, Ida and Bahr, Frauke S and Krüger, Anna-Lena and Labus, Josephine and Singh, Ghanendra and Ponimaskin, Evgeni G and Lenk, Kerstin and Zeug, Andre},
title = {{LPS-induced inflammation differentially affects endogenous Ca\&lt;sup\&gt;2\&lt;/sup\&gt;⁺ activity in mouse and human iPSC-derived astrocytes}},
journal = {Molecular medicine (Cambridge, Mass.)},
year = {2026},
month = mar,
volume = {32},
number = {1},
pages = {52},
publisher = {The Feinstein Institute for Medical Research},
issn = {1076-1551},
doi = {10.1186/s10020-026-01450-3},
url = {https://doi.org/10.1186/s10020-026-01450-3},
pmid = {41787269},
pmcid = {PMC13064076}
}

RIS

TY - JOUR
AU - Müller, Franziska E
AU - Ivanov, Flavian
AU - Studt, Anne-Catharine
AU - Nitzsche, Ida
AU - Bahr, Frauke S
AU - Krüger, Anna-Lena
AU - Labus, Josephine
AU - Singh, Ghanendra
AU - Ponimaskin, Evgeni G
AU - Lenk, Kerstin
AU - Zeug, Andre
TI - LPS-induced inflammation differentially affects endogenous Ca&lt;sup&gt;2&lt;/sup&gt;⁺ activity in mouse and human iPSC-derived astrocytes
T2 - Molecular medicine (Cambridge, Mass.)
J2 - Mol Med
PY - 2026
DA - 2026/03/05
VL - 32
IS - 1
SP - 52
SN - 1076-1551
PB - The Feinstein Institute for Medical Research
DO - 10.1186/s10020-026-01450-3
UR - https://doi.org/10.1186/s10020-026-01450-3
LA - en
ER -

CSL-JSON

{
"id": "10.1186/s10020-026-01450-3",
"type": "article-journal",
"title": "LPS-induced inflammation differentially affects endogenous Ca&lt;sup&gt;2&lt;/sup&gt;⁺ activity in mouse and human iPSC-derived astrocytes",
"container-title": "Molecular medicine (Cambridge, Mass.)",
"author": [
{
"family": "Müller",
"given": "Franziska E"
},
{
"family": "Ivanov",
"given": "Flavian"
},
{
"family": "Studt",
"given": "Anne-Catharine"
},
{
"family": "Nitzsche",
"given": "Ida"
},
{
"family": "Bahr",
"given": "Frauke S"
},
{
"family": "Krüger",
"given": "Anna-Lena"
},
{
"family": "Labus",
"given": "Josephine"
},
{
"family": "Singh",
"given": "Ghanendra"
},
{
"family": "Ponimaskin",
"given": "Evgeni G"
},
{
"family": "Lenk",
"given": "Kerstin"
},
{
"family": "Zeug",
"given": "Andre"
}
],
"container-title-short": "Mol Med",
"volume": "32",
"issue": "1",
"page": "52",
"DOI": "10.1186/s10020-026-01450-3",
"PMID": "41787269",
"PMCID": "PMC13064076",
"ISSN": "1076-1551",
"publisher": "The Feinstein Institute for Medical Research",
"URL": "https://doi.org/10.1186/s10020-026-01450-3",
"language": "en",
"issued": {
"date-parts": [
[
2026,
3,
5
]
]
}
}

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:10.1111/jnc.70470 [code]
Calcium Buffering in Astrocytes and Its Relevance for Experimental Data Interpretation and Computational Modeling.
Journal: Journal of neurochemistry
In common: cellular / molecular, 5 references, 2 authors
[2] doi:10.1186/s12974-026-03837-9
PTP1B in astrocytes drives pathogen-induced neurodegeneration.
Journal: Journal of neuroinflammation
In common: mouse, cellular / molecular, 6 references
[3] doi:10.1038/s41467-026-73435-z [code]
Cellular and subcellular heterogeneity of astrocytic Na⁺ homeostasis tuning astrocytes into functionally distinct subgroups in the mouse brain.
Journal: Nature communications
In common: seaborn, pandas, Matplotlib, 1 other tool, mouse, cellular / molecular, 3 references
[4] doi:10.3390/ijms27135902
Functional Expression of Nicotinic Receptors on iPSC-Derived Astrocytes and Signalling Disturbances by a Panel of Neonicotinoid Pesticides and Their Metabolites.
Journal: International journal of molecular sciences
In common: cellular / molecular, 5 references
[5] doi:10.1167/iovs.67.5.73 [code]
Identification of a Small-Molecule Modulator of Astrocyte Reactivity for Optic Nerve Protection.
Journal: Investigative ophthalmology & visual science
In common: scikit-learn, pandas, Matplotlib, 1 other tool, mouse, cellular / molecular, 3 references
[6] doi:10.1002/alz.71745 [code]
Regional astrocyte dysregulation and altered glymphatic-related markers in Alzheimer's disease frontal cortex.
Journal: Alzheimer's & dementia : the journal of the Alzheimer's Association
In common: seaborn, pandas, Matplotlib, 1 other tool, cellular / molecular, 3 references
[7] doi:10.1111/ejn.70480 [code]
Astrocyte Proximity Protects Synapses From Human Amyloid-Beta Induced Degeneration in a Mouse Ex Vivo Model of Early Alzheimer's Disease.
Journal: The European journal of neuroscience
In common: seaborn, scikit-learn, pandas, 2 other tools, mouse, cellular / molecular, 2 references
[8] doi:10.1016/j.celrep.2026.117803
APOE3 and APOE4 human astrocytes differentially modulate Alzheimer's disease pathology and microglial responses in chimeric mice.
Journal: Cell reports
In common: mouse, cellular / molecular, 4 references
[9] doi:10.1038/s41514-026-00391-9 [code]
Region-specific transcriptional signatures of brain aging in the absence of neuropathology at the single-cell level.
Journal: npj aging
In common: seaborn, scikit-learn, pandas, 2 other tools, cellular / molecular, 2 references
[10] doi:10.1038/s41467-026-76837-1 [code]
Drug screen and machine learning predict neuroprotective agents in a preclinical human model of childhood dementia.
Journal: Nature communications
In common: seaborn, scikit-learn, pandas, 2 other tools, 2 references

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.