Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 51 of the 51 references it cites were found in a registry or an index.

A new Bayesian factor analysis method improves detection of genes and biological processes affected by perturbations in single-cell CRISPR screening

Open access

Abstract

Clustered regularly interspaced short palindromic repeats (CRISPR) screening coupled with single-cell RNA sequencing has emerged as a powerful tool to characterize the effects of genetic perturbations on the whole transcriptome at a single-cell level. However, due to its sparsity and complex structure, analysis of single-cell CRISPR screening data is challenging. In particular, standard differential expression analysis methods are often underpowered to detect genes affected by CRISPR perturbations. We developed a statistical method for such data, called guided sparse factor analysis (GSFA). GSFA infers latent factors that represent coregulated genes or gene modules; by borrowing information from these factors, it infers the effects of genetic perturbations on individual genes. We demonstrated through extensive simulation studies that GSFA detects perturbation effects with much higher power than state-of-the-art methods. Using single-cell CRISPR data from human CD8+ T cells and neural progenitor cells, we showed that GSFA identified biologically relevant gene modules and specific genes affected by CRISPR perturbations, many of which were missed by existing methods, providing new insights into the functions of genes involved in T cell activation and neurodevelopment.

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 · 10.1038/s41592-023-02017-4

Cites

51 of 51 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 GSFA model and its application on real data. Top, the input of the GSFA includes the perturbation matrix and the gene expression matrix. Bottom, the output of GSFA includes the effects of perturbations on targets (β), the gene loading matrix (W) and the list of genes affected by each perturbation after LFSR thresholding. The box shows how the GSFA calculates the total effect of a perturbation on the expression of individual genes.
Fig. 1 GSFA model and its application on real data. Top, the input of the GSFA includes the perturbation matrix and the gene expression matrix. Bottom, the output of GSFA includes the effects of perturbations on targets (β), the gene loading matrix (W) and the list of genes affected by each perturbation after LFSR thresholding. The box shows how the GSFA calculates the total effect of a perturbation on the expression of individual genes.CC BY · PubMed Central
Fig. 2 GSFA performance on simulated data. a, Distributions of the absolute correlation values between true factors and the factors inferred by GSFA under the normal setting. The different colors represent different values of true factor density varying from 0.05 to 0.2. b, Same as in a but under count-based scenarios. c, Box plots of absolute effect sizes from perturbation factor regression estimated by GSFA under the normal setting. The different colors represent different values of true factor density varying from 0.05 to 0.2. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper and lower whiskers extend from the hinge to the largest and smallest values no further than 1.5× the interquartile range from the hinge. d, Same as in c but under count-based scenarios. e, Receiver operating characteristic (ROC) curves of DEG discovery under the count-based setting and three different levels of true factor density; the four colors correspond to four DEG detection methods. The results shown are of perturbations with a true association effect of 0.3 on factors. Each curve was a mean representation over 300 datasets generated under the corresponding setting, with the mean area under the curve (AUC) labeled in colored text. See Supplementary Figs. 1 and 2 for results under other settings. f, Distributions of the observed FDPs among significant DEGs detected using GSFA (LFSR < 0.05) and other methods (FDR < 0.05) per dataset under the count-based setting and several true factor densities. The four colors correspond to four DEG detection methods. Source data
Fig. 2 GSFA performance on simulated data. a, Distributions of the absolute correlation values between true factors and the factors inferred by GSFA under the normal setting. The different colors represent different values of true factor density varying from 0.05 to 0.2. b, Same as in a but under count-based scenarios. c, Box plots of absolute effect sizes from perturbation factor regression estimated by GSFA under the normal setting. The different colors represent different values of true factor density varying from 0.05 to 0.2. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper and lower whiskers extend from the hinge to the largest and smallest values no further than 1.5× the interquartile range from the hinge. d, Same as in c but under count-based scenarios. e, Receiver operating characteristic (ROC) curves of DEG discovery under the count-based setting and three different levels of true factor density; the four colors correspond to four DEG detection methods. The results shown are of perturbations with a true association effect of 0.3 on factors. Each curve was a mean representation over 300 datasets generated under the corresponding setting, with the mean area under the curve (AUC) labeled in colored text. See Supplementary Figs. 1 and 2 for results under other settings. f, Distributions of the observed FDPs among significant DEGs detected using GSFA (LFSR < 0.05) and other methods (FDR < 0.05) per dataset under the count-based setting and several true factor densities. The four colors correspond to four DEG detection methods. Source dataCC BY · PubMed Central
Fig. 3 GSFA results of inferred factors from the analysis of CROP-seq data of primary CD8+ T cells. The results are based on stimulated T cells. a, Estimated effects of gene perturbations on selected factors inferred by GSFA. The size of a dot represents the PIP of the association; the color represents the effect size. b, Venn diagram of targets identified using the original clustering-based method versus GSFA. c, Loading of selected marker genes on selected factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d, Fold enrichment of selected GO ‘biological process’ gene sets significantly enriched (q < 0.05) in factors 2, 4, 9 and 12. Each bar is colored by −log10 P values from the overrepresentation test (an upper-tailed hypergeometric test), where overlap of a gene set with genes with a PIP > 0.95 in the factor was compared against that of all genes used in the GSFA. Source data
Fig. 3 GSFA results of inferred factors from the analysis of CROP-seq data of primary CD8+ T cells. The results are based on stimulated T cells. a, Estimated effects of gene perturbations on selected factors inferred by GSFA. The size of a dot represents the PIP of the association; the color represents the effect size. b, Venn diagram of targets identified using the original clustering-based method versus GSFA. c, Loading of selected marker genes on selected factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d, Fold enrichment of selected GO ‘biological process’ gene sets significantly enriched (q < 0.05) in factors 2, 4, 9 and 12. Each bar is colored by −log10 P values from the overrepresentation test (an upper-tailed hypergeometric test), where overlap of a gene set with genes with a PIP > 0.95 in the factor was compared against that of all genes used in the GSFA. Source dataCC BY · PubMed Central
Fig. 4 GSFA results of the effects of genetic perturbation on gene expression in CD8+ T cell data. Results are based on stimulated CD8+ T cells. a, Number of DEGs detected under all perturbations using four different methods. The y axis is log-scaled and the bar height corresponds to count +1 (as the number of DEGs could be 0); the exact numbers of DEGs are labeled on top of the bars. The detection threshold for DEGs is LFSR < 0.05 for GSFA and FDR < 0.05 for all other methods. b, Number of GO Slim ‘biological process’ terms enriched in DEGs detected using different methods. c, Heatmap of selected GO ‘biological process’ terms and their fold enrichment in DEGs (LFSR < 0.05) detected using GSFA under different perturbations. d, GSFA estimated the effects of perturbations on marker genes in stimulated T cells. The sizes of the dots represent LFSR bins; the colors of the dots represent the summarized effect sizes. e, scMAGeCK estimated effects of perturbations on marker genes in stimulated T cells. The sizes of the dots represent the FDR bins; the colors of the dots represent the scMAGeCK selection scores. f, A target–marker–phenotype regulatory network summarizing the GSFA results. Significant (LFSR < 0.05) regulatory relationships between target and marker genes are represented by the colored arrows, with the red sharp arrows indicating positive regulation of marker genes by the target genes, and the blue blunt arrows indicating negative regulation. The darkness of the color represents the relative effect magnitude. Note that the effect directions here are the opposite of the perturbation effects. Source data
Fig. 4 GSFA results of the effects of genetic perturbation on gene expression in CD8+ T cell data. Results are based on stimulated CD8+ T cells. a, Number of DEGs detected under all perturbations using four different methods. The y axis is log-scaled and the bar height corresponds to count +1 (as the number of DEGs could be 0); the exact numbers of DEGs are labeled on top of the bars. The detection threshold for DEGs is LFSR < 0.05 for GSFA and FDR < 0.05 for all other methods. b, Number of GO Slim ‘biological process’ terms enriched in DEGs detected using different methods. c, Heatmap of selected GO ‘biological process’ terms and their fold enrichment in DEGs (LFSR < 0.05) detected using GSFA under different perturbations. d, GSFA estimated the effects of perturbations on marker genes in stimulated T cells. The sizes of the dots represent LFSR bins; the colors of the dots represent the summarized effect sizes. e, scMAGeCK estimated effects of perturbations on marker genes in stimulated T cells. The sizes of the dots represent the FDR bins; the colors of the dots represent the scMAGeCK selection scores. f, A target–marker–phenotype regulatory network summarizing the GSFA results. Significant (LFSR < 0.05) regulatory relationships between target and marker genes are represented by the colored arrows, with the red sharp arrows indicating positive regulation of marker genes by the target genes, and the blue blunt arrows indicating negative regulation. The darkness of the color represents the relative effect magnitude. Note that the effect directions here are the opposite of the perturbation effects. Source dataCC BY · PubMed Central
Fig. 5 GSFA analysis of the CROP-seq data of LUHMES cells. a, Estimated effects of gene perturbations on selected factors inferred using GSFA. The size of a dot represents the PIP of association; the color represents the effect size. b, Venn diagram of targets identified from the original pseudotime association analysis versus from the GSFA. c, Loading of neuronal marker genes on selected factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d, Fold of enrichment of selected GO ‘biological process’ terms enriched in factors 4, 9 and 16 (q < 0.05). Each bar is colored using −log10 P values from the overrepresentation test (an upper-tailed hypergeometric test), where overlap of a gene set with genes with PIP > 0.95 in the factor was compared against that of all genes used in the GSFA. e, Number of DEGs detected under all perturbations using four different methods. The y axis is log-scaled and the bar height corresponds to count +1 (as the number of DEGs could be 0); the exact number of DEGs is labeled above the bars. The detection threshold for DEGs is LFSR < 0.05 for GSFA and FDR < 0.05 for all other methods. f, Number of GO Slim ‘biological process’ terms enriched in DEGs detected using different methods. g, GSFA estimated effects of perturbations on marker genes. The sizes of the dots represent the LFSR bins; the colors of the dots represent the summarized effect sizes. h, scMAGeCK estimated effects of perturbations on marker genes. The sizes of the dots represent the FDR bins; the colors of the dots represent the scMAGeCK selection scores. i, Target–marker–phenotype regulatory network summarizing the GSFA results. Significant (LFSR < 0.05) regulatory relationships between target and marker genes are represented by the colored arrows, with the red sharp arrows indicating positive regulation of marker genes by target genes, and the blue blunt arrows indicating negative regulation. The darkness of the color represents the relative magnitude of effect. Note that the direction of regulation is the opposite of the perturbation effect. Source data
Fig. 5 GSFA analysis of the CROP-seq data of LUHMES cells. a, Estimated effects of gene perturbations on selected factors inferred using GSFA. The size of a dot represents the PIP of association; the color represents the effect size. b, Venn diagram of targets identified from the original pseudotime association analysis versus from the GSFA. c, Loading of neuronal marker genes on selected factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d, Fold of enrichment of selected GO ‘biological process’ terms enriched in factors 4, 9 and 16 (q < 0.05). Each bar is colored using −log10 P values from the overrepresentation test (an upper-tailed hypergeometric test), where overlap of a gene set with genes with PIP > 0.95 in the factor was compared against that of all genes used in the GSFA. e, Number of DEGs detected under all perturbations using four different methods. The y axis is log-scaled and the bar height corresponds to count +1 (as the number of DEGs could be 0); the exact number of DEGs is labeled above the bars. The detection threshold for DEGs is LFSR < 0.05 for GSFA and FDR < 0.05 for all other methods. f, Number of GO Slim ‘biological process’ terms enriched in DEGs detected using different methods. g, GSFA estimated effects of perturbations on marker genes. The sizes of the dots represent the LFSR bins; the colors of the dots represent the summarized effect sizes. h, scMAGeCK estimated effects of perturbations on marker genes. The sizes of the dots represent the FDR bins; the colors of the dots represent the scMAGeCK selection scores. i, Target–marker–phenotype regulatory network summarizing the GSFA results. Significant (LFSR < 0.05) regulatory relationships between target and marker genes are represented by the colored arrows, with the red sharp arrows indicating positive regulation of marker genes by target genes, and the blue blunt arrows indicating negative regulation. The darkness of the color represents the relative magnitude of effect. Note that the direction of regulation is the opposite of the perturbation effect. Source dataCC BY · PubMed Central
Extended Data Fig. 1 Additional GSFA results on simulated data. a) Comparison of estimated factor densities using two priors under the count-based setting. b) The proportion of truly associated factor-gene pairs out of all the pairs that have GSFA estimated gene loading PIP > 0.95 in the corresponding factor, computed for each dataset under three levels of true factor density and the normal setting. c) Same as in b) but under the count-based setting. d) The proportion of truly associated perturbation-factor pairs out of all the pairs that have GSFA estimated association PIP > 0.95, computed for each dataset under three levels of true factor density and the normal setting. e) Same as in d) but under the count-based setting. For each box in b), c), d) and e), n = 300 proportion values generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. f) Observed proportion of false discoveries among significant DEGs detected by GSFA (LFSR < 0.05) or Welch’s t-test (FDR < 0.05), computed for each dataset under three levels of true factor density and the normal setting. Source data
Extended Data Fig. 1 Additional GSFA results on simulated data. a) Comparison of estimated factor densities using two priors under the count-based setting. b) The proportion of truly associated factor-gene pairs out of all the pairs that have GSFA estimated gene loading PIP > 0.95 in the corresponding factor, computed for each dataset under three levels of true factor density and the normal setting. c) Same as in b) but under the count-based setting. d) The proportion of truly associated perturbation-factor pairs out of all the pairs that have GSFA estimated association PIP > 0.95, computed for each dataset under three levels of true factor density and the normal setting. e) Same as in d) but under the count-based setting. For each box in b), c), d) and e), n = 300 proportion values generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. f) Observed proportion of false discoveries among significant DEGs detected by GSFA (LFSR < 0.05) or Welch’s t-test (FDR < 0.05), computed for each dataset under three levels of true factor density and the normal setting. Source dataCC BY · PubMed Central
Extended Data Fig. 2 Simulation results of two-step clustering analysis vs. GSFA. Panel a) shows Normal based simulation. The first row of Panel a) shows the false positive rates of the discovered DEGs across different factor density settings. The second row shows the power of detecting associations of a guide with a cluster or factor. The clustering method here is based on K-means. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. Panel b) Same as in a) but under the count-based setting. Clustering analysis was done using Seurat. Source data
Extended Data Fig. 2 Simulation results of two-step clustering analysis vs. GSFA. Panel a) shows Normal based simulation. The first row of Panel a) shows the false positive rates of the discovered DEGs across different factor density settings. The second row shows the power of detecting associations of a guide with a cluster or factor. The clustering method here is based on K-means. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. Panel b) Same as in a) but under the count-based setting. Clustering analysis was done using Seurat. Source dataCC BY · PubMed Central
Extended Data Fig. 3 Simulation results under the setting where gRNAs have non-specific effects (see Methods). a, b) Estimation of beta under the normal scenario (a) and the count-based scenario (b). M1-M6, perturbations; M7, negative control. The true effect sizes, after adjusting for negative control should be 0 for all except M5. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. c) Histogram of the proportions of false discoveries observed among DEGs found by GSFA at LFSR < 0.05, across 300 simulations. The results of a simple t-test comparing targeted cells against negative control cells are included for comparison, with DEGs discovered at FDR < 0.05. Source data
Extended Data Fig. 3 Simulation results under the setting where gRNAs have non-specific effects (see Methods). a, b) Estimation of beta under the normal scenario (a) and the count-based scenario (b). M1-M6, perturbations; M7, negative control. The true effect sizes, after adjusting for negative control should be 0 for all except M5. For each box, n = 300 estimates generated from 300 rounds of simulation under the given setting; the center line of the box represents the median; the lower and upper hinges of the box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. c) Histogram of the proportions of false discoveries observed among DEGs found by GSFA at LFSR < 0.05, across 300 simulations. The results of a simple t-test comparing targeted cells against negative control cells are included for comparison, with DEGs discovered at FDR < 0.05. Source dataCC BY · PubMed Central
Extended Data Fig. 4 GSFA vs. two-step factor analysis in simulations. Shown are the false discovery proportions (FDP) of the DEGs detected by either method. Source data
Extended Data Fig. 4 GSFA vs. two-step factor analysis in simulations. Shown are the false discovery proportions (FDP) of the DEGs detected by either method. Source dataCC BY · PubMed Central
Extended Data Fig. 5 The performance of GSFA compared with MAST and Wilcoxon on the simulation dataset, where target genes of perturbation were chosen randomly. Each panel in the figure displays the results for three gRNAs with varying effect sizes, measured by standard deviations. The differential analysis is comparing each gRNA against cells perturbed by negative control gRNA. Panel a) shows the false discovery proportion (FDP) of GSFA, MAST, and Wilcoxon with effect size being 0.3, 0.4, and 0.5. For each box, n = 50 estimates generated from 50 rounds of simulation under the given setting; The centerline of a box represents the median; the lower and upper hinges of a box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. Panel b) shows the ROC of these three methods. Source data
Extended Data Fig. 5 The performance of GSFA compared with MAST and Wilcoxon on the simulation dataset, where target genes of perturbation were chosen randomly. Each panel in the figure displays the results for three gRNAs with varying effect sizes, measured by standard deviations. The differential analysis is comparing each gRNA against cells perturbed by negative control gRNA. Panel a) shows the false discovery proportion (FDP) of GSFA, MAST, and Wilcoxon with effect size being 0.3, 0.4, and 0.5. For each box, n = 50 estimates generated from 50 rounds of simulation under the given setting; The centerline of a box represents the median; the lower and upper hinges of a box correspond to the first and third quartiles; the upper/lower whisker extends from the hinge to the largest/smallest value no further than 1.5 * inter-quartile range from the hinge. Panel b) shows the ROC of these three methods. Source dataCC BY · PubMed Central
Extended Data Fig. 6 Additional GSFA results on CD8+ T cell CROP-seq dataset. a) Estimated effects of gene perturbations on all factors inferred by GSFA within stimulated T cells. The size of a dot represents the PIP of association; the color represents the effect size. b) Similar to a) but estimated within unstimulated T cells. c) Loading of selected marker genes on all factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d–f) Estimated effects of perturbations on marker genes in stimulated T cells with DESeq2 (d), MAST (e), and SCEPTRE (f). Sizes of the dots represent FDR bins; colors of the dots represent the DESeq2 log2 fold change estimates, the MAST log fold change estimates, and the SCEPTER log fold change estimates, respectively. Source data
Extended Data Fig. 6 Additional GSFA results on CD8+ T cell CROP-seq dataset. a) Estimated effects of gene perturbations on all factors inferred by GSFA within stimulated T cells. The size of a dot represents the PIP of association; the color represents the effect size. b) Similar to a) but estimated within unstimulated T cells. c) Loading of selected marker genes on all factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. d–f) Estimated effects of perturbations on marker genes in stimulated T cells with DESeq2 (d), MAST (e), and SCEPTRE (f). Sizes of the dots represent FDR bins; colors of the dots represent the DESeq2 log2 fold change estimates, the MAST log fold change estimates, and the SCEPTER log fold change estimates, respectively. Source dataCC BY · PubMed Central
Extended Data Fig. 7 Assessing GSFA results for differential expression (DE) analysis using MAST. a), b) Comparison of DEGs found by GSFA vs. MAST. a) T cells. b) LUHMES. c), d) Quantile-quantile plot of p-values of differential expression estimated by MAST of GSFA detected DE genes, assuming a uniform(0,1) null distribution. c) T cells. d) LUHMES. Source data
Extended Data Fig. 7 Assessing GSFA results for differential expression (DE) analysis using MAST. a), b) Comparison of DEGs found by GSFA vs. MAST. a) T cells. b) LUHMES. c), d) Quantile-quantile plot of p-values of differential expression estimated by MAST of GSFA detected DE genes, assuming a uniform(0,1) null distribution. c) T cells. d) LUHMES. Source dataCC BY · PubMed Central
Extended Data Fig. 8 Additional GSFA results on LUHMES CROP-seq dataset. a) Estimated effects of gene perturbations on all factors inferred by GSFA. The size of a dot represents the PIP of association; the color represents the effect size. b) Loading of neuronal marker genes on all factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. c) Heatmap of selected GO ‘biological process’ terms and their folds of enrichment in DEGs detected by GSFA (LFSR < 0.05). d–f) Estimated effects of perturbations on marker genes in LUHMES with DESeq2 (d), MAST (e), and SCEPTRE (f). Sizes of the dots represent FDR bins; colors of the dots represent the DESeq2 log2 fold change estimates, the MAST log fold change estimates, and the SCEPTER log fold change estimates, respectively. Source data
Extended Data Fig. 8 Additional GSFA results on LUHMES CROP-seq dataset. a) Estimated effects of gene perturbations on all factors inferred by GSFA. The size of a dot represents the PIP of association; the color represents the effect size. b) Loading of neuronal marker genes on all factors. The size of a dot represents the gene PIP in a factor and the color represents the gene weight (magnitude of contribution) in a factor. c) Heatmap of selected GO ‘biological process’ terms and their folds of enrichment in DEGs detected by GSFA (LFSR < 0.05). d–f) Estimated effects of perturbations on marker genes in LUHMES with DESeq2 (d), MAST (e), and SCEPTRE (f). Sizes of the dots represent FDR bins; colors of the dots represent the DESeq2 log2 fold change estimates, the MAST log fold change estimates, and the SCEPTER log fold change estimates, respectively. Source dataCC 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 50 references

Research works printed but not linked: 2

  1. Welch, B. L. The generalisation of student’s problems when several different population variances are involved. Biometrika 34, 28–35 (1947).
  2. O’Hara, R. B. & Sillanpää, M. J. A review of Bayesian variable selection methods: what, how and which. Bayesian Anal. 4, 85–117 (2009).

Cited by

See all 34

Metadata

Authors
Zhou, Y., Luo, K., Liang, L., Chen, M., & He, X.
Affiliations
University of Chicago (US)
Corresponding author
  • Chen, Mengjie
  • He, Xin
Venue
Nature Methods
Published
2023-09-28
Volume
20
Issue
11
Pages
1693-1703
ISSN
  • 1548-7091 (ISSN-L)
  • 1548-7105
OpenAlex
W4387130415
Cited by
34
References
50
Citations by year
2026: 8 · 2025: 12 · 2024: 13 · 2023: 1
FWCI
3.03
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Methods (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Knowledge@UChicago (University of Chicago) (submittedVersion, cc-by-nc-nd)
  • University of Chicago (cc-by)
Funders
  • U.S. Department of Health & Human Services | National Institutes of Health (R01MH110531, R01MH116281, R01HG010773, R01GM126553, R01HG011883)
  • National Science Foundation (2016307)
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
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › CRISPR and Genetic Engineering
Keywords
Bayesian factor analysis, differential expression detection, gene modules, co-regulated genes, single-cell RNA sequencing, CD8+ T cells, neural progenitor cells, T cell activation, brain development
MeSH terms
Bayes Theorem, Biological Phenomena, Humans, CD8-Positive T-Lymphocytes, Transcriptome, Clustered Regularly Interspaced Short Palindromic Repeats, CRISPR-Cas Systems

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: A new Bayesian factor analysis method improves detection of genes and biological processes affected by perturbations in single-cell CRISPR screening. https://app.cochat.ai/papers/10.1038/s41592-023-02017-4
The paper
Zhou, Y., Luo, K., Liang, L., Chen, M., & He, X. (2023). A new Bayesian factor analysis method improves detection of genes and biological processes affected by perturbations in single-cell CRISPR screening. Nature Methods. https://doi.org/10.1038/s41592-023-02017-4

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.