OSCR

Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data.

Code ↔ Paper

3 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 3 matches
  1. [1] § Methods › Data processing › TCGA pan-gynecologic atlas ↔ sofa/utils/utils.py, lines 18–105 · score 0.60 · highly variable features, log transformed, variance
  2. [2] § Methods › Data processing › Pan-cancer DepMap ↔ sofa/utils/utils.py, lines 18–105 · score 0.59 · highly variable genes, log transformed
  3. [3] § Methods › Downstream analysis › Gene set overrepresentation analysis (ORA) ↔ sofa/plots/plots.py, lines 348–407 · score 0.52 · enrichr API, overrepresentation, background, gene

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

Python · 504 lines · 16 KB · MIT · 2 matches

  1. #!/usr/bin/env python3
  2. import pyro
  3. import torch
  4. import numpy as np
  5. import muon as mu
  6. from muon import MuData
  7. from sklearn.preprocessing import LabelEncoder
  8. from anndata import AnnData
  9. from ..models.SOFA import SOFA
  10. import pandas as pd
  11. import scanpy as sc
  12. from typing import Union
  13. import numpy as np
  14. from sklearn.preprocessing import LabelEncoder, StandardScaler
  15. import gseapy as gp
  16. from sklearn.metrics import log_loss, accuracy_score
  18. def get_ad(data: pd.DataFrame,
  19. llh: str="gaussian",
  20. select_hvg: bool=False,
  21. log: bool=False,
  22. scale: bool=False,
  23. scaling_factor: float=0.1
  24. ) -> AnnData:
  25. """
  26. Convert a numpy array to an AnnData object.
  28. Parameters:
  29. -----------
  30. data : pandas DataFrame
  31. The input data to be converted to AnnData object.
  32. name : str
  33. The name of the variable.
  34. llh : str, optional
  35. The likelihood of the data. It should be "gaussian", "bernoulli" or
  36. "categorical". Default is "gaussian".
  37. select_hvg: bool, optional
  38. whether to select highly variable features.
  39. log: bool, optional
  40. whether to log transform the data.
  41. scale: bool, optional
  42. whether to center and scale the data.
  43. scaling_factor: float, optional
  44. The scaling factor to scale the likelihood for this view.
  45. It is a crucial hyperparameter. Too high scaling parameters can lead to overfitting
  46. (Factors don't explain variance of Xmdata, but perfectly predict Ymdata).
  47. Too low values can lead to the model ignoring the covariate guidance. In practice
  48. a value of 0.1 is a good starting point. Default is 0.1.
  49. Returns:
  50. --------
  51. adata : AnnData
  52. The converted AnnData object.
  53. """
  54. data =data.loc[:,~data.columns.duplicated()]
  55. data_ = data.loc[~np.all(pd.isnull(data), axis=1),:]
  56. data = data.loc[:,~np.any(pd.isnull(data_), axis=0)]
  57. mask = ~np.any(pd.isnull(data), axis=1)
  58. mask.index = mask.index.astype(str)
  59. if llh == "multinomial" or llh == "bernoulli":
  60. if type(data) != int:
  61. label_encoder = LabelEncoder()
  62. # Apply LabelEncoder to each column
  63. encoded_data = data.values.flatten()
  64. encoded_data = label_encoder.fit_transform(encoded_data)
  65. encoded_data = encoded_data.reshape(data.shape)
  66. adata = AnnData(encoded_data, dtype=np.float32)
  67. label_mapping = {label: encoded_label for label, encoded_label in zip(label_encoder.classes_, label_encoder.transform(label_encoder.classes_))}
  68. else:
  69. encoded_data = data.values
  71. if data.shape[1] == 1:
  72. adata = AnnData(encoded_data.reshape(-1,1), dtype=np.float32)
  73. adata.var_names = data.columns
  74. else:
  75. adata = AnnData(encoded_data, dtype=np.float32)
  76. adata.var_names = data.columns
  77. data.index = data.index.astype(str)
  78. adata.obs_names = data.index.tolist()
  79. if log:
  80. sc.pp.log1p(adata)
  81. adata.obsm["mask"] = mask.values
  83. if select_hvg:
  84. adata_filtered = adata[~np.all(data.isna(),axis=1),:]
  86. adata = adata[:,~np.any(np.isnan(adata_filtered.X),axis=0)]
  88. adata_filtered = adata_filtered[:,~np.any(np.isnan(adata_filtered.X),axis=0)]
  90. sc.pp.highly_variable_genes(adata_filtered, n_top_genes=2000)
  91. adata = adata[:,adata_filtered.var["highly_variable"]]
  92. adata.var["highly_variable"] = adata_filtered.var["highly_variable"]
  93. if scale:
  94. scaler = StandardScaler()
  95. adata.X = scaler.fit_transform(adata.X)
  97. adata.X[adata.obsm["mask"] == False] = 0
  98. adata.uns["llh"] = llh
  101. adata.uns["scaling_factor"] = scaling_factor
  105. return adata
  107. def calc_var_explained(X_pred, X):
  108. """
  109. Calculate R2 for X and X_pred.
  111. Parameters
  112. ----------
  113. X_pred : numpy.array
  114. Predicted X.
  115. X : numpy.array
  116. Input X.
  118. Returns
  119. -------
  120. float
  121. R2 value for X and X_pred.
  122. """
  123. num = np.sum(np.square(X-X_pred))
  124. denom = np.sum(np.square(X))
  125. vexp= 1 - num/denom
  126. if vexp < 0:
  127. vexp = 0
  128. return vexp
  132. def get_var_explained_per_view_factor(model: SOFA):
  133. """
  134. Calculate the fraction of variance of each view
  135. that is explained by each factor.
  136. Parameters
  137. ----------
  138. model : SOFA
  139. The trained SOFA model.
  140. Returns
  141. -------
  142. numpy.array
  143. Array containing the fraction of variance of each view
  144. that is explained by each factor.
  145. """
  146. X = [i.cpu().numpy() for i in model.X]
  147. vexp = []
  148. if not hasattr(model, "Z"):
  149. model.Z = model.predict("Z", num_split=10000)
  150. if not hasattr(model, f"W"):
  151. model.W = [model.predict(f"W_{i}", num_split=10000) for i in range(len(X))]
  152. for i in range(len(X)):
  153. mask = model.Xmask[i].cpu().numpy()
  154. vexp_factor = []
  155. for j in range(model.num_factors):
  156. X_pred_factor = model.Z[mask,j, np.newaxis] @ model.W[i][np.newaxis,j,:]
  157. vexp_factor.append(calc_var_explained(X_pred_factor, X[i][mask,:]))
  158. vexp.append(np.stack(vexp_factor).reshape(model.num_factors,1))
  159. vexp = np.hstack(vexp)
  160. return vexp
  162. def calc_var_explained_(X_pred, X):
  163. """
  164. Calculate the fraction of variance of each view
  165. that is explained by each factor.
  167. Parameters
  168. ----------
  169. X_pred : numpy.array
  170. Predicted X.
  171. X : numpy.array
  172. Input X.
  174. Returns
  175. -------
  176. numpy.array
  177. Array containing the fraction of variance of each view
  178. that is explained by each factor.
  179. """
  180. vexp = []
  181. for i in range(len(X_pred)):
  182. num = np.sum(np.square(X-X_pred[i]))
  183. denom = np.sum(np.square(X))
  184. vexp.append(1 - num/denom)
  185. vexp = np.stack(vexp)
  186. vexp[vexp < 0] = 0
  187. return vexp
  191. def get_loadings(model: SOFA,
  192. view: str
  193. )-> pd.DataFrame:
  194. """
  195. Get the loadings of the model for a specific view.
  197. Parameters
  198. ----------
  199. model : SOFA
  200. The trained SOFA model.
  201. view : str
  202. Name of the view to get the loadings for.
  204. Returns
  205. -------
  206. pd.DataFrame
  207. DataFrame containing the loadings of the model for the specified view.
  208. """
  209. ind_labels = np.array([f"Factor_{i+1}" for i in range(model.num_factors)], dtype=object)
  210. if model.Ymdata is not None:
  211. guided_factors = list(model.Ymdata.mod.keys())
  212. for i in range(len(guided_factors)):
  213. s = " (" + guided_factors[i] + ")"
  214. ind_labels[model.design.cpu().numpy()[i,:]==1] = ind_labels[model.design.cpu().numpy()[i,:]==1] + s
  216. if hasattr(model, f"W"):
  217. W = pd.DataFrame(model.W[model.views.index(view)], index = ind_labels, columns = model.Xmdata.mod[view].var_names)
  218. else:
  219. model.W = [model.predict(f"W_{i}", num_split=10000) for i in range(len(model.X))]
  220. W = pd.DataFrame(model.W[model.views.index(view)], index = ind_labels, columns = model.Xmdata.mod[view].var_names)
  221. return W
  223. def get_factors(model: SOFA,
  224. )-> pd.DataFrame:
  225. """
  226. Get the loadings of the model for a specific view.
  228. Parameters
  229. ----------
  230. model : SOFA
  231. The trained SOFA model.
  232. Returns
  233. -------
  234. pd.DataFrame
  235. DataFrame containing the loadings of the model for the specified view.
  236. """
  238. col_labels = np.array([f"Factor_{i+1}" for i in range(model.num_factors)], dtype=object)
  239. if model.Ymdata is not None:
  240. guided_factors = list(model.Ymdata.mod.keys())
  241. for i in range(len(guided_factors)):
  242. s = " (" + guided_factors[i] + ")"
  243. col_labels[model.design.cpu().numpy()[i,:]==1] = col_labels[model.design.cpu().numpy()[i,:]==1] + s
  245. if hasattr(model, f"Z"):
  246. Z = pd.DataFrame(model.Z, index = model.Xmdata.obs.index, columns = col_labels)
  247. else:
  248. model.Z = model.predict("Z")
  249. Z = pd.DataFrame(model.Z, index = model.Xmdata.obs.index, columns = col_labels)
  250. return Z
  252. def get_top_loadings(model,view, factor, sign="+", top_n=100):
  253. """
  254. Get the top_n loadings of the model for a specific view.
  256. Parameters
  257. ----------
  258. model : SOFA
  259. The trained SOFA model.
  260. view : str
  261. Name of the view to get the loadings for.
  262. factor : int
  263. Index of the factor to get the top loadings for.
  264. Should be between 1 and the total number of factors.
  265. sign : str
  266. Sign of the loadings to get. Default is "+".
  267. top_n : int
  268. Number of top loadings to get. Default is 100.
  269. Returns
  270. -------
  271. pandas.DataFrame
  272. DataFrame containing the top_n loadings of the model for the specified view.
  273. """
  274. assert(sign=="+" or sign=="-")
  275. # correct for pythonic indexing
  276. factor = factor-1
  277. W = get_loadings(model, view)
  278. W = W.iloc[factor,:]
  279. if sign == "+":
  280. idx = np.argpartition(W, -top_n)[-top_n:]
  281. topW = W.index[idx]
  282. elif sign=="-":
  283. idx = np.argpartition(W*-1, -top_n)[-top_n:]
  284. topW = W.index[idx]
  285. return topW.tolist()
  287. def get_gsea_enrichment(gene_list, db, background):
  288. """
  289. Get gene set enrichment analysis results based on a gene_list using gseapy.
  291. Parameters
  292. ----------
  293. gene_list : list
  294. List of strings containing gene names.
  295. db : list
  296. List of strings containing database names to be used for enrichment analysis.
  297. background : list
  298. List of strings containing gene names to be used as background.
  299. Returns
  300. -------
  301. Enrichr object
  302. Enrichr object containing the results of the enrichment analysis.
  303. """
  304. enr = gp.enrichr(gene_list=gene_list, # or "./tests/data/gene_list.txt",
  305. gene_sets=[db],
  306. organism='human', # don't forget to set organism to the one you desired! e.g. Yeast
  307. outdir=None,# don't write to disk
  308. background=background
  309. )
  310. return enr
  312. def calc_rmse(X, X_pred):
  313. """
  314. Calculate the root mean squared error between X and X_pred.
  316. Parameters
  317. ----------
  318. X_pred : numpy.array
  319. Predicted X.
  320. X : numpy.array
  321. Input X.
  323. Returns
  324. -------
  325. float
  326. The root mean squared error of X of the model.
  327. """
  328. rmse = np.sqrt(np.sum(np.square(X-X_pred))/(X.shape[0]*X.shape[1]))
  329. return rmse
  331. def get_rmse(model):
  332. """
  333. Calculate the root mean squared error of the model.
  335. Parameters
  336. ----------
  337. model : SOFA
  338. THe trained SOFA model.
  340. Returns
  341. -------
  342. dict
  343. The root mean squared error of X of the model for each view.
  344. """
  345. if not hasattr(model, f"X_pred"):
  346. model.X_pred = [model.predict(f"X_{i}", num_split=10000) for i in range(len(model.X))]
  347. rmse = {}
  348. for i in range(len(model.X)):
  349. rmse[model.views[i]] = calc_rmse(model.X[i].cpu().numpy(), model.X_pred[i])
  350. return rmse
  352. def sigmoid(x):
  353. return 1 / (1 + np.exp(-x))
  355. def softmax(x):
  356. return np.exp(x) / np.sum(np.exp(x))
  358. def get_guide_error(model):
  359. """
  360. Calculate the root mean squared error for continuous, binary crossentropy for binary or
  361. categorical cross entropy for categorical Y of the model.
  363. Parameters
  364. ----------
  365. model : SOFA
  366. The trained SOFA model.
  368. Returns
  369. -------
  370. dict
  371. Containing the root mean squared error for continuous, binary crossentropy for binary or
  372. categorical cross entropy for categorical Y of the model.
  373. """
  374. if model.Ymdata is None:
  375. raise ValueError("Model does not have guide variables!")
  376. if not hasattr(model, "Y_pred"):
  377. model.Y_pred = []
  378. for i in range(len(model.Y)):
  379. model.Y_pred.append(model.predict(f"Y_{i}"))
  381. error = {}
  383. for ix, (i,j) in enumerate(zip(model.Y, model.Y_pred)):
  384. i = i.cpu().numpy()
  385. if model.guide_llh[ix] == "gaussian":
  386. error[model.guide_views[ix]]= calc_rmse(i,j)
  387. elif model.guide_llh[ix] == "bernoulli":
  388. error[model.guide_views[ix]] = log_loss(i,sigmoid(j))
  389. elif model.guide_llh[ix] == "multinomial":
  390. error[model.guide_views[ix]] = log_loss(i,softmax(j))
  391. return error
  394. def save_model(model, file_prefix):
  395. """
  396. Saves a model as h5mu and save files to disk.
  397. Model hyperparameters, input data and predictions are saved in the h5mu file
  398. and the model parameters are saved in the save file.
  399. Both files are needed to load a model and continue training.
  401. Parameters
  402. ----------
  403. model : SOFA
  404. The trained SOFA model.
  405. file_prefix : str
  406. Filename prefix to save the model as h5mu and save files.
  407. Returns
  408. -------
  409. tuple(str,str)
  410. Filenames of the saved h5mu and save files.
  411. """
  413. model_mdata = model.save_as_mudata()
  414. try:
  415. model_mdata.write(file_prefix+".h5mu")
  416. except RuntimeError:
  417. print("Metadata probably too long to save. Please save the metadata separately. \nSaving model without metadata!")
  418. # remove metadata if saving does not work
  419. model_mdata.obs = pd.DataFrame(index=model_mdata.obs.index)
  420. # save model without metadata
  421. model_mdata.write(file_prefix+".h5mu")
  422. except TypeError:
  423. print("Mixed type columns are currently not supported when saving metadata of a model. Please save the metadata separately.\nSaving model without metadata!")
  424. # remove metadata if saving does not work
  425. model_mdata.obs = pd.DataFrame(index=model_mdata.obs.index)
  426. # save model without metadata
  427. model_mdata.write(file_prefix+".h5mu")
  428. except Exception:
  429. print("Unexpected error! \nSaving model without metadata!")
  430. # remove metadata if saving does not work
  431. model_mdata.obs = pd.DataFrame(index=model_mdata.obs.index)
  432. # save model without metadata
  433. model_mdata.write(file_prefix+".h5mu")
  436. dict_ = pyro.get_param_store()
  437. dict_.save(file_prefix +".save")
  438. return file_prefix +".h5mu", file_prefix +".save"
  442. def load_model(file_prefix):
  443. """
  444. Load a saved model from disk.
  445. The function requires an h5mu and a save file to load model.
  447. Parameters
  448. ----------
  449. file_prefix : str
  450. Filename prefix to save the model as h5mu and save files.
  452. Returns
  453. -------
  454. SOFA
  455. The loaded SOFA model.
  456. """
  457. mdata = mu.read(file_prefix +".h5mu")
  458. if "guide_mod" in list(mdata.uns.keys()):
  459. Ymdata = MuData({i:mdata.mod[i] for i in mdata.uns["guide_mod"]})
  460. Xmdata = MuData({i:mdata.mod[i] for i in mdata.mod if i not in mdata.uns["guide_mod"]})
  461. design = mdata.uns["input_design"]
  462. else:
  463. Xmdata = MuData({i:mdata.mod[i] for i in mdata.mod})
  464. Ymdata = None
  465. design = np.array(0)
  466. num_factors = mdata.uns["input_num_factors"]
  467. # TODO find way to save and load mixed column metadata
  468. # check if entries in obs (metadata)
  469. if mdata.obs.shape[1] !=0:
  470. metadata = mdata.obs
  471. else:
  472. metadata = None
  474. horseshoe = mdata.uns["horseshoe"]
  475. # apparently mu.read does not read None type in uns so
  476. # need to check if seed is there, if not set to None
  477. if "seed" not in mdata.uns:
  478. seed = None
  479. else:
  480. seed = mdata.uns["seed"]
  481. pyro.set_rng_seed(seed)
  482. device = mdata.uns["device"]
  483. model = SOFA(Xmdata,
  484. num_factors=num_factors,
  485. Ymdata = Ymdata,
  486. design = torch.tensor(design),
  487. device=device,
  488. horseshoe=horseshoe,
  489. subsample=0,
  490. metadata = metadata,
  491. seed=seed)
  492. model.Z = mdata.uns["Z"]
  493. W = [mdata.uns[f"W_{i}"] for i in model.views]
  494. model.W = W
  495. model.X_pred =[mdata.uns[f"X_{i}"] for i in model.views]
  496. if Ymdata is not None:
  497. model.Y_pred = [mdata.uns[f"Y_pred_{i}"] for i in Ymdata.mod]
  499. model.history = mdata.uns["history"]
  501. # load pyro paramstore
  502. dict_ = pyro.get_param_store()
  503. dict_.load(file_prefix+".save")
  504. return model

utils.py at commit 2e56def, under MIT · at the source

Overview

Authors: Tümay Capraz1,2, Harald Vöhringer2,3,4,5, Klaus Sebastian Augusto Kruger Serrano6, Ricardo Omar Ramirez Flores7, Julio Saez-Rodriguez6,7, Wolfgang Huber2,3
ORCID iDs: Tümay Capraz, Harald Vöhringer, Klaus Sebastian Augusto Kruger Serrano, Ricardo Omar Ramirez Flores, Julio Saez-Rodriguez, Wolfgang Huber
  1. Collaboration for joint PhD degree between EMBL and Heidelberg University, Faculty of Biosciences,Heidelberg, Germany
  2. Genome Biology Unit, EMBL,Heidelberg, Germany
  3. Molecular Medicine Partnership Unit (MMPU), Heidelberg, Germany
  4. Department of Medicine V, Hematology, Oncology and Rheumatology, University Hospital Heidelberg,Heidelberg, Germany
  5. Department of Hematology and Oncology, University Hospital Düsseldorf,Düsseldorf, Germany
  6. Heidelberg University, Faculty of Medicine, and Heidelberg University Hospital, Institute for Computational Biomedicine,Heidelberg, Germany
  7. European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI),Hinxton, UK
Institutions: Heidelberg University (Germany); European Molecular Biology Laboratory (Germany); University Hospital Heidelberg (Germany); Düsseldorf University Hospital (Germany); European Bioinformatics Institute (United Kingdom)
Journal: Nature communications, volume 17, issue 1, article 8725
Dates: received 7 March 2025; accepted 11 June 2026; published online 20 August 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1038/s41467-026-74694-6 · PMID 42624833 · PMCID PMC13493367 · OpenAlex W6893097162
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: genetics / omics (modality), human (organism), methods / tools (subfield)
Methods: Statistics, Smoothing, state filtering, decompositions, Connectivity
Keywords: Data integration, Software, Statistical methods
MeSH: Multiomics*, Brain, Factor Analysis, Statistical, Heart Failure, Humans, Neoplasms, Proteomics (* major topic)
Topic: Advanced Proteomics Techniques and Applications (Spectroscopy, Chemistry), according to OpenAlex
Funding: This work was funded by the European Research Council (Synergy Grant DECODE under grant agreement no. 810296)
Citations: not cited yet (Europe PMC); 95 references in the paper

Abstract

A fundamental design pattern in biomolecular studies is to assay the same set of samples (organisms, tissue biopsies, or individual cells) by multiple different ‘omics assays. Group Factor Analysis (GFA) and its adaptation to high-dimensional settings, Multi-Omics Factor Analysis (MOFA), are widely used as a first-line approach to analyze such data and are effective in detecting patterns of correlation, organize them into so-called latent factors, and identify common and assay-specific factors. However, in many applications, a subset of the found factors just rediscovers already known covariates (e.g., disease subtypes, environmental covariates) while others may represent genuine novelty.

Here, we present Semi-supervised Omics Factor Analysis (SOFA), a method that incorporates known covariates into the model upfront and focuses the factor discovery on novel sources of variation. We show SOFA’s effectiveness for discovering novel patterns by applying it to cancer, brain development and heart failure multi-omic data sets.

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

Repositories

Its files are read in the Code ↔ Paper reader above, with 3 matches between paragraphs and lines of code.

Zenodo 14761127

License: CC-BY-4.0
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Size: 1 file
Software Heritage: not checked
Found in: “Code availability”
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
At the source:

tcapraz/SOFA

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 2e56defd2ef8ced64bf642c121297712b5f8b8bc, 15 October 2025
Languages: Python (11), Jupyter (4)
Size: 30 files, 15 scripts
Software Heritage: not archived
Found in: “Code availability”
Holds: README, license file, environment (pyproject.toml, docs/requirements.txt), tests, continuous integration, documentation, 4 notebooks
Not found: CITATION.cff
Tools: NumPy (9 files), pandas (8 files), PyTorch (8 files), anndata (6 files), scikit-learn (5 files), Matplotlib (4 files), Scanpy (4 files), seaborn (4 files), Pyro (3 files), statsmodels (2 files), SciPy (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
17 files
At the source: github.com/tcapraz/SOFA

Zenodo 19594991

License: CC-BY-4.0
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Size: 1 file
Software Heritage: not checked
Found in: “Code availability”
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
At the source: zenodo.org/records/19594991

Code availability

SOFA is implemented in the Python package biosofa and is available on github (https://github.com/tcapraz/SOFA) and PyPI (https://pypi.org/project/biosofa/), along with documentation and example analysis notebooks. Code and processed data to reproduce the analysis and figures are available on Zenodo 10.5281/zenodo.14761127. The package version used to perform the presented analyses and generate the figures are available on Zenodo https://zenodo.org/records/19594991.

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

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:

  • 3 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 15 scripts, each with its path and the digest of its content;
  • 3 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

Processed data to reproduce the analysis and figures are available on Zenodo 10.5281/zenodo.14761127.Source data are provided with this paper.

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, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 6 authors, 3 keywords, 7 MeSH terms, 1 funder, 82 references.

Cite

This paper

Capraz, T., Vöhringer, H., Kruger Serrano, K. S. A., Ramirez Flores, R. O., Saez-Rodriguez, J., & Huber, W. (2026). Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data. Nature communications, 17(1), 8725. https://doi.org/10.1038/s41467-026-74694-6

BibTeX

@article{capraz2026semi,
author = {Capraz, Tümay and Vöhringer, Harald and Kruger Serrano, Klaus Sebastian Augusto and Ramirez Flores, Ricardo Omar and Saez-Rodriguez, Julio and Huber, Wolfgang},
title = {{Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data}},
journal = {Nature communications},
year = {2026},
month = aug,
volume = {17},
number = {1},
pages = {8725},
publisher = {Nature Publishing Group},
issn = {2041-1723},
doi = {10.1038/s41467-026-74694-6},
url = {https://doi.org/10.1038/s41467-026-74694-6},
pmid = {42624833},
pmcid = {PMC13493367}
}

RIS

TY - JOUR
AU - Capraz, Tümay
AU - Vöhringer, Harald
AU - Kruger Serrano, Klaus Sebastian Augusto
AU - Ramirez Flores, Ricardo Omar
AU - Saez-Rodriguez, Julio
AU - Huber, Wolfgang
TI - Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data
T2 - Nature communications
J2 - Nat Commun
PY - 2026
DA - 2026/08/20
VL - 17
IS - 1
SP - 8725
SN - 2041-1723
PB - Nature Publishing Group
DO - 10.1038/s41467-026-74694-6
UR - https://doi.org/10.1038/s41467-026-74694-6
LA - en
ER -

CSL-JSON

{
"id": "10.1038/s41467-026-74694-6",
"type": "article-journal",
"title": "Semi-supervised Omics Factor Analysis (SOFA) disentangles known and latent sources of variation in multi-omic data",
"container-title": "Nature communications",
"author": [
{
"family": "Capraz",
"given": "Tümay"
},
{
"family": "Vöhringer",
"given": "Harald"
},
{
"family": "Kruger Serrano",
"given": "Klaus Sebastian Augusto"
},
{
"family": "Ramirez Flores",
"given": "Ricardo Omar"
},
{
"family": "Saez-Rodriguez",
"given": "Julio"
},
{
"family": "Huber",
"given": "Wolfgang"
}
],
"container-title-short": "Nat Commun",
"volume": "17",
"issue": "1",
"page": "8725",
"DOI": "10.1038/s41467-026-74694-6",
"PMID": "42624833",
"PMCID": "PMC13493367",
"ISSN": "2041-1723",
"publisher": "Nature Publishing Group",
"URL": "https://doi.org/10.1038/s41467-026-74694-6",
"language": "en",
"issued": {
"date-parts": [
[
2026,
8,
20
]
]
}
}

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:10.1093/bioinformatics/btag652 [code]
mmVelo: a deep generative model for estimating cell state-dependent dynamics across multiple modalities.
Journal: Bioinformatics (Oxford, England)
In common: Pyro, anndata, Scanpy, 8 other tools, genetics / omics, 5 references
[2] doi:10.1038/s44320-026-00208-7 [code]
Interpretable deep generative ensemble learning for single-cell omics with Hydra.
Journal: Molecular systems biology
In common: anndata, Scanpy, PyTorch, 6 other tools, 5 references
[3] doi:10.1016/j.celrep.2026.117110 [code]
Single-nucleus multiome analysis in the human prefrontal cortex identifies gene expression and cis-regulatory elements associated with aging.
Journal: Cell reports
In common: anndata, Scanpy, statsmodels, 6 other tools, genetics / omics, 3 references
[4] doi:10.1523/eneuro.0362-25.2026 [code]
Similarities between &lt;i&gt;Ciona&lt;/i&gt; Dorsal Motor Ganglion and Vertebrate Cerebellum: Did a Chordate Ancestor Already Show D/V Subdivision within a Hindbrain Precursor?
Journal: eNeuro
In common: Pyro, anndata, Scanpy, 8 other tools, 1 reference
[5] doi:10.1038/s41592-026-03057-2 [code]
CREsted: modeling genomic and synthetic cell-type-specific enhancers across tissues and species.
Journal: Nature methods
In common: anndata, Scanpy, statsmodels, 7 other tools, methods / tools, genetics / omics, 2 references
[6] 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: anndata, Scanpy, statsmodels, 7 other tools, genetics / omics, 2 references
[7] doi:10.1016/j.xgen.2026.101217 [code]
ProtoCloud: A prototypical self-explaining model for single-cell analysis.
Journal: Cell genomics
In common: anndata, Scanpy, PyTorch, 6 other tools, genetics / omics, 3 references
[8] doi:10.64898/2026.03.30.714220 [code]
An integrated single cell and spatial omics atlas of human prenatal development
Journal: bioRxiv (preprint)
In common: anndata, Scanpy, statsmodels, 7 other tools, 2 references
[9] doi:10.1093/bib/bbag298 [code]
Empowering multifaceted analysis of spatial transcriptomics data with RGAST.
Journal: Briefings in bioinformatics
In common: anndata, Scanpy, PyTorch, 6 other tools, methods / tools, genetics / omics, 2 references
[10] doi:10.1002/advs.77003 [code]
SemanticST: A Scalable Multi-Contextual Graph Learning Framework for Uncovering Spatial Niches and Robust Multi-Sample Integration in Spatial Transcriptomics.
Journal: Advanced science (Weinheim, Baden-Wurttemberg, Germany)
In common: anndata, Scanpy, PyTorch, 6 other tools, methods / tools, genetics / omics, 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.