CoChat

Sfaira accelerates data and model reuse in single cell genomics

Open access

Published version of a preprint

The DOI matches its Crossref record; 77 of the 77 references it cites were found in a registry or an index.

Abstract

Single-cell RNA-seq datasets are often first analyzed independently without harnessing model fits from previous studies, and are then contextualized with public data sets, requiring time-consuming data wrangling. We address these issues with sfaira, a single-cell data zoo for public data sets paired with a model zoo for executable pre-trained models. The data zoo is designed to facilitate contribution of data sets using ontologies for metadata. We propose an adaption of cross-entropy loss for cell type classification tailored to datasets annotated at different levels of coarseness. We demonstrate the utility of sfaira by training models across anatomic data partitions on 8 million cells.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/s13059-021-02452-6

Cites

77 of 77 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Sfaira is a data and model zoo that automates common steps in exploratory single-cell RNA-seq analysis. a Overview workflow of sfaira data and model API. Data set files are currently stored in cloud databases that can be interfaced with the sfaira data API to give streamlined AnnData objects that can be used for analysis or model fitting. The sfaira model API can consume these data sets to produce automatised analyses by querying parameter estimates from pre-trained models stored on cloud servers via the sfaira parameter API. Example analysis steps that are automated are embedding computation and cell type annotation. b Summary of the current state of sfaira data zoo for mouse and human single-cell RNA-seq samples, representing 220 data sets and 8.0 million cells in two organisms
Fig. 1 Sfaira is a data and model zoo that automates common steps in exploratory single-cell RNA-seq analysis. a Overview workflow of sfaira data and model API. Data set files are currently stored in cloud databases that can be interfaced with the sfaira data API to give streamlined AnnData objects that can be used for analysis or model fitting. The sfaira model API can consume these data sets to produce automatised analyses by querying parameter estimates from pre-trained models stored on cloud servers via the sfaira parameter API. Example analysis steps that are automated are embedding computation and cell type annotation. b Summary of the current state of sfaira data zoo for mouse and human single-cell RNA-seq samples, representing 220 data sets and 8.0 million cells in two organismsCC BY · PubMed Central
Fig. 2 The sfaira data zoo contextualizes data statistics. a Characterizing gene expression distributions across organs. The expression range of the example gene insulin across mouse data sets shows specific expression in pancreas. Mean normalized expression of Ins1 by organ and cell type. b Pearson correlation coefficient between the two cell-cycle-associated genes Pcna and Mcm5 across data sets. Shown is a boxplot of the distribution over the correlation coefficients for each data set computed per organ and cell type. c Data sets vary strongly in complexity. Shown is the number of cells versus the number of cell types in the data zoo by organ for both mouse and human. d Sfaira allows querying of cell type fractions in tissue across organisms. Shown is the fraction of T cells per mouse dataset, ordered by organ. e Mean total counts per cell in mouse and human organs for 10x protocol data sets only. f Number of data sets per experimental protocol. g Mean number of counts (unique molecular identifiers if available, otherwise reads) per data set by experimental protocol
Fig. 2 The sfaira data zoo contextualizes data statistics. a Characterizing gene expression distributions across organs. The expression range of the example gene insulin across mouse data sets shows specific expression in pancreas. Mean normalized expression of Ins1 by organ and cell type. b Pearson correlation coefficient between the two cell-cycle-associated genes Pcna and Mcm5 across data sets. Shown is a boxplot of the distribution over the correlation coefficients for each data set computed per organ and cell type. c Data sets vary strongly in complexity. Shown is the number of cells versus the number of cell types in the data zoo by organ for both mouse and human. d Sfaira allows querying of cell type fractions in tissue across organisms. Shown is the fraction of T cells per mouse dataset, ordered by organ. e Mean total counts per cell in mouse and human organs for 10x protocol data sets only. f Number of data sets per experimental protocol. g Mean number of counts (unique molecular identifiers if available, otherwise reads) per data set by experimental protocolCC BY · PubMed Central
Fig. 3 Sfaira automatizes exploratory analysis of single-cell data. a Manual single-cell data analysis pipeline and automated sfaira pipeline. b Comparison of manual feature engineering workflow with automated embedding and cell type annotation from sfaira on a human PBMC data set. Shown are UMAPs based on a PCA of an engineered feature space and of the out-of-the-box latent space from a linear sfaira embedding model. Superimposed are cell types previously annotated for this data set and sfaira cell type predictions
Fig. 3 Sfaira automatizes exploratory analysis of single-cell data. a Manual single-cell data analysis pipeline and automated sfaira pipeline. b Comparison of manual feature engineering workflow with automated embedding and cell type annotation from sfaira on a human PBMC data set. Shown are UMAPs based on a PCA of an engineered feature space and of the out-of-the-box latent space from a linear sfaira embedding model. Superimposed are cell types previously annotated for this data set and sfaira cell type predictionsCC BY · PubMed Central
Fig. 4 Sfaira allows fitting of cell type classifiers for data sets with different levels of annotation granularity by using cell type ontologies. a Aggregated accuracy and cross-entropy allow for fitting cell type classification models on data sets with heterogeneous annotation coarsity using cell type relations from ontologies (see the “Methods” section). The y axis contains leaf nodes of a cell type ontology, which can be combined linearly to yield the predicted probability mass of any other node in the ontology graph (x-axis). b, c Accuracy of cell type classifiers on mouse (b) and human (c) organs on entirely held-out test-data sets. Linear: Linear classifier (logistic regression), marker: Marker gene-based classifier, MLP: multilayer dense neural network. d, e Class-wise prediction accuracy correlates with the number of cells in class. Shown are cell type class-wise F1 scores by the number of cell types in that class of cell type classifiers by model on lung data from mice (d) and humans (e)
Fig. 4 Sfaira allows fitting of cell type classifiers for data sets with different levels of annotation granularity by using cell type ontologies. a Aggregated accuracy and cross-entropy allow for fitting cell type classification models on data sets with heterogeneous annotation coarsity using cell type relations from ontologies (see the “Methods” section). The y axis contains leaf nodes of a cell type ontology, which can be combined linearly to yield the predicted probability mass of any other node in the ontology graph (x-axis). b, c Accuracy of cell type classifiers on mouse (b) and human (c) organs on entirely held-out test-data sets. Linear: Linear classifier (logistic regression), marker: Marker gene-based classifier, MLP: multilayer dense neural network. d, e Class-wise prediction accuracy correlates with the number of cells in class. Shown are cell type class-wise F1 scores by the number of cell types in that class of cell type classifiers by model on lung data from mice (d) and humans (e)CC BY · PubMed Central
Fig. 5 Sfaira allows streamlined embedding models training across tissues and on whole atlases. a, b Pre-trained embedding models can perform meaningful reconstruction of cells in held out data sets. UMAP based on latent space of the best embedding model data for pancreas data from humans (a) and mice (b). The superimposed colors correspond to the original, non-streamlined, cell type annotation. c, d Reconstruction performance comparison of different embedding models across organs and organisms. The negative binomial likelihood is used as a reconstruction performance metric on reconstructed test data of held-out test-data sets from PCA, linear, non-negative matrix factorization (nmf), autoencoder (ae), and variational autoencoder (vae) models on human (c) and mouse (d) organs. e, f Sfaira allows for training of embedding models using very large data sets. UMAP of the latent space of an embedding model trained on all mouse data in the sfaira data zoo with the data set (e) and cell type (f) superimposed
Fig. 5 Sfaira allows streamlined embedding models training across tissues and on whole atlases. a, b Pre-trained embedding models can perform meaningful reconstruction of cells in held out data sets. UMAP based on latent space of the best embedding model data for pancreas data from humans (a) and mice (b). The superimposed colors correspond to the original, non-streamlined, cell type annotation. c, d Reconstruction performance comparison of different embedding models across organs and organisms. The negative binomial likelihood is used as a reconstruction performance metric on reconstructed test data of held-out test-data sets from PCA, linear, non-negative matrix factorization (nmf), autoencoder (ae), and variational autoencoder (vae) models on human (c) and mouse (d) organs. e, f Sfaira allows for training of embedding models using very large data sets. UMAP of the latent space of an embedding model trained on all mouse data in the sfaira data zoo with the data set (e) and cell type (f) superimposedCC BY · PubMed Central
Fig. 6 Toward the interpretability of model embedding. Saliency-based interpretation and data regularization of non-conditional embedding models: linear and autoencoder embedding models for human esophagus (a, b). a Correlation of cell-type wise aggregated gradients of embedding with respect to input features. b Distribution of feature-wise wise aggregated gradients of embedding with respect to input features by cell type (color)
Fig. 6 Toward the interpretability of model embedding. Saliency-based interpretation and data regularization of non-conditional embedding models: linear and autoencoder embedding models for human esophagus (a, b). a Correlation of cell-type wise aggregated gradients of embedding with respect to input features. b Distribution of feature-wise wise aggregated gradients of embedding with respect to input features by cell type (color)CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 77 references

Research works printed but not linked: 3

  1. Dony L, Koenig M, Fischer D, Theis FJ. Variational autoencoders with flexible priors enable robust distribution learning on single-cell RNA sequencing data. In: ICML 2020 Workshop on Computational Biology (WCB) Proceedings Paper 37; 2020.
  2. Fischer DS, Dony L, König M, Moeed A, Zappia L, Heumos L, et al. Theis. PyPi: Sfaira accelerates data and model reuse in single cell genomics. PyPi https://pypi.org/project/sfaira/0.3.0/. 2021. · DOI
  3. Fischer DS, Dony L, König M, Moeed A, Zappia L, Heumos L, et al. Theis. Zenodo: Sfaira accelerates data and model reuse in single cell genomics. Zenodo https://zenodo.org/record/4836517. 2021. · DOI

Cited by

See all 33

Metadata

Authors
Fischer, D. S., Dony, L., König, M., Moeed, A., Zappia, L., Heumos, L., Tritschler, S., Holmberg, O., Aliee, H., & Theis, F. J.
Affiliations
Helmholtz Munich (DE)
Corresponding author
Theis, Fabian J.
Venue
Genome Biology
Published
2021-08-25
Volume
22
Issue
1
Pages
248-248
ISSN
  • 1465-6906 (ISSN-L)
  • 1474-760X
OpenAlex
W3194078460
Cited by
33
References
75
Citations by year
2026: 1 · 2025: 5 · 2024: 5 · 2023: 11 · 2022: 6 · 2021: 5
FWCI
2.12
Type
journal article
Language
en
Open access
Licence
Locations
  • Genome biology (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PuSH - Publication Server of Helmholtz Zentrum München (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • mediaTUM (Technical University of Munich) (submittedVersion, other-oa)
  • Site cant be reached (submittedVersion)
  • Site cant be reached (submittedVersion, cc-by)
Funders
  • Deutsche Forschungsgemeinschaft (GSC 1006)
  • Bundesministerium für Bildung und Forschung (01IS18036B, 01IS18053A)
  • Chan Zuckerberg Initiative (2019-207271)
  • Helmholtz Zentrum München (ZT-I-PF-5-01)
  • Joachim Herz Stiftung (-)
  • Helmholtz Zentrum München - Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
  • Social Sciences › Decision Sciences › Information Systems and Management › Scientific Computing and Data Management
Keywords
single-cell RNA sequencing, data reuse, model reuse, pre-trained models, model zoo, metadata ontology, cross-entropy loss, cell type classification, multi-omics integration
MeSH terms
Animals, Humans, Statistics as Topic, Reproducibility of Results, Genomics, Databases, Genetic, Mice, Molecular Sequence Annotation, Single-Cell Analysis, Gene Ontology

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Sfaira accelerates data and model reuse in single cell genomics. https://app.cochat.ai/papers/10.1186/s13059-021-02452-6
The paper
Fischer, D. S., Dony, L., König, M., Moeed, A., Zappia, L., Heumos, L., Tritschler, S., Holmberg, O., Aliee, H., & Theis, F. J. (2021). Sfaira accelerates data and model reuse in single cell genomics. Genome Biology. https://doi.org/10.1186/s13059-021-02452-6

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.