CoChat

The DOI matches its Crossref record; 33 of the 34 references it cites were found in a registry or an index.

SCEPTRE improves calibration and sensitivity in single-cell CRISPR screen analysis

Open access

Published version of a preprint

Abstract

Single-cell CRISPR screens are a promising biotechnology for mapping regulatory elements to target genes at genome-wide scale. However, technical factors like sequencing depth impact not only expression measurement but also perturbation detection, creating a confounding effect. We demonstrate on two single-cell CRISPR screens how these challenges cause calibration issues. We propose SCEPTRE: analysis of single-cell perturbation screens via conditional resampling, which infers associations between perturbations and expression by resampling the former according to a working model for perturbation detection probability in each cell. SCEPTRE demonstrates very good calibration and sensitivity on CRISPR screen data, yielding hundreds of new regulatory relationships supported by orthogonal biological evidence.

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/s13059-021-02545-2

Cites

33 of 33 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 CRISPR screen analysis challenges can lead to false positives and false negatives. a QQ-plot of negative control p-values produced by Gasperini et al. (red; downsampled for visualization) and Xie et al. (gray-green). These p-values deviate substantially from the expected uniform distribution, indicating test miscalibration. b–d Technical factors, such as sequencing depth and batch, impact gRNA detection probability and observed gene expression levels in both Gasperini et al. (b) and Xie et al. (c) data. Thus, technical factors act as confounders (d), differentiating CRISPR screens from traditional differential expression applications. e Monocle2 estimates the dispersion of each gene by projecting each gene’s raw dispersion estimate onto the fitted raw dispersion-mean expression curve. This estimation procedure leads to miscalibration for high-dispersion genes
Fig. 1 CRISPR screen analysis challenges can lead to false positives and false negatives. a QQ-plot of negative control p-values produced by Gasperini et al. (red; downsampled for visualization) and Xie et al. (gray-green). These p-values deviate substantially from the expected uniform distribution, indicating test miscalibration. b–d Technical factors, such as sequencing depth and batch, impact gRNA detection probability and observed gene expression levels in both Gasperini et al. (b) and Xie et al. (c) data. Thus, technical factors act as confounders (d), differentiating CRISPR screens from traditional differential expression applications. e Monocle2 estimates the dispersion of each gene by projecting each gene’s raw dispersion estimate onto the fitted raw dispersion-mean expression curve. This estimation procedure leads to miscalibration for high-dispersion genesCC BY · PubMed Central
Fig. 2 SCEPTRE: Analysis of single-cell perturbation screens via conditional resampling. A schematic and outline of the SCEPTRE methodology for one gene and one gRNA. SCEPTRE estimates the probability of gRNA detection in each cell based on its technical factors. It then builds a null distribution for the negative binomial z-value by independently resampling gRNA presence or absence for each cell according to these probabilities to form “negative control” datasets. A skew-t distribution is fit to the resulting histogram to obtain precise p-values based on a limited number of resamples, against which the original NB z-value is compared. The dashed line shows the standard normal distribution, against which the NB z-value typically would be compared
Fig. 2 SCEPTRE: Analysis of single-cell perturbation screens via conditional resampling. A schematic and outline of the SCEPTRE methodology for one gene and one gRNA. SCEPTRE estimates the probability of gRNA detection in each cell based on its technical factors. It then builds a null distribution for the negative binomial z-value by independently resampling gRNA presence or absence for each cell according to these probabilities to form “negative control” datasets. A skew-t distribution is fit to the resulting histogram to obtain precise p-values based on a limited number of resamples, against which the original NB z-value is compared. The dashed line shows the standard normal distribution, against which the NB z-value typically would be comparedCC BY · PubMed Central
Fig. 3 SCEPTRE demonstrates good calibration and sensitivity under known ground truth. a Numerical simulation comparing SCEPTRE and improved negative binomial regression on four simulated datasets. The negative binomial model was incorrectly specified on three of the four datasets. SCEPTRE maintained good calibration across all four datasets, despite model misspecification and confounder presence. b, c Application of SCEPTRE, improved negative binomial regression, Monocle regression, and Virtual FACS to pairs of negative control gRNAs and genes in b the Gasperini et al. data and c the Xie et al. data. Compared to the other methods, SCEPTRE showed good calibration. d SCEPTRE p-values for Gasperini et al. TSS-targeting controls were highly significant, and in general, more significant than the original empirical p-values. e Comparison of p-values produced by SCEPTRE for ARL15-enh to p-values produced by an arrayed, bulk RNA-seq CRISPR screen of ARL15-enh. The results of the two analyses coincided almost exactly, with both analyses rejecting gene ARL15 with high confidence after a BH correction. Dotted blue lines, rejection thresholds
Fig. 3 SCEPTRE demonstrates good calibration and sensitivity under known ground truth. a Numerical simulation comparing SCEPTRE and improved negative binomial regression on four simulated datasets. The negative binomial model was incorrectly specified on three of the four datasets. SCEPTRE maintained good calibration across all four datasets, despite model misspecification and confounder presence. b, c Application of SCEPTRE, improved negative binomial regression, Monocle regression, and Virtual FACS to pairs of negative control gRNAs and genes in b the Gasperini et al. data and c the Xie et al. data. Compared to the other methods, SCEPTRE showed good calibration. d SCEPTRE p-values for Gasperini et al. TSS-targeting controls were highly significant, and in general, more significant than the original empirical p-values. e Comparison of p-values produced by SCEPTRE for ARL15-enh to p-values produced by an arrayed, bulk RNA-seq CRISPR screen of ARL15-enh. The results of the two analyses coincided almost exactly, with both analyses rejecting gene ARL15 with high confidence after a BH correction. Dotted blue lines, rejection thresholdsCC BY · PubMed Central
Fig. 4 Application of SCEPTRE to Gasperini et al. data yields biologically plausible gene-enhancer links. a Comparison of the original empirical p-values to those obtained from SCEPTRE. The two analysis methods differed substantially, with 200 gene-enhancer links discovered only by SCEPTRE and 107 discovered only by the original analysis. Annotations correspond to pairs in panel (e). b Distribution of distances from TSS to upstream paired enhancers. Compared to Monocle NB (original) and improved NB analyses, SCEPTRE paired genes with nearer enhancers on average. c For those gene-enhancer pairs falling in the same TAD, the cumulative distribution of the fractional rank of the HI-C interaction frequency compared to other distance-matched loci pairs within the same TAD. SCEPTRE showed similar enrichment despite finding 25% more within-TAD pairs compared to the original analysis. Inset table shows gene-enhancer pairs falling in the same TAD. SCEPTRE found 93 more total pairs compared to the original analysis, and a higher percentage of pairs fell within the same TAD. d Enrichment of ChIP-seq signal from seven cell-type relevant transcription factors and one histone mark (H3K27ac) among paired enhancers. SCEPTRE showed stronger enrichment across all ChIP-seq targets. e Five gene-enhancer pairs discovered by SCEPTRE but not the original analysis, each supported by a whole blood GTEx eQTL or FANTOM enhancer RNA correlation p-value
Fig. 4 Application of SCEPTRE to Gasperini et al. data yields biologically plausible gene-enhancer links. a Comparison of the original empirical p-values to those obtained from SCEPTRE. The two analysis methods differed substantially, with 200 gene-enhancer links discovered only by SCEPTRE and 107 discovered only by the original analysis. Annotations correspond to pairs in panel (e). b Distribution of distances from TSS to upstream paired enhancers. Compared to Monocle NB (original) and improved NB analyses, SCEPTRE paired genes with nearer enhancers on average. c For those gene-enhancer pairs falling in the same TAD, the cumulative distribution of the fractional rank of the HI-C interaction frequency compared to other distance-matched loci pairs within the same TAD. SCEPTRE showed similar enrichment despite finding 25% more within-TAD pairs compared to the original analysis. Inset table shows gene-enhancer pairs falling in the same TAD. SCEPTRE found 93 more total pairs compared to the original analysis, and a higher percentage of pairs fell within the same TAD. d Enrichment of ChIP-seq signal from seven cell-type relevant transcription factors and one histone mark (H3K27ac) among paired enhancers. SCEPTRE showed stronger enrichment across all ChIP-seq targets. e Five gene-enhancer pairs discovered by SCEPTRE but not the original analysis, each supported by a whole blood GTEx eQTL or FANTOM enhancer RNA correlation p-valueCC BY · PubMed Central
Fig. 5 SCEPTRE discovers biologically plausible gene-enhancer links on Xie et al. data. a Comparison of SCEPTRE p-values to Virtual FACS significance scores. Significant SCEPTRE p-values (n=139) are colored in blue and purple, and the top 139 Virtual FACS pairs, as ranked by significance score, are colored in gray-green and purple. The two sets diverged substantially, with only 43% of pairs in shared across sets. b–d These panels are similar to the corresponding panels in Fig. 4. SCEPTRE pairs showed strong enrichment for biological signals associated with enhancer activity on b physical distance, c HIC interaction, and d ChIP-seq metrics relative to other methods
Fig. 5 SCEPTRE discovers biologically plausible gene-enhancer links on Xie et al. data. a Comparison of SCEPTRE p-values to Virtual FACS significance scores. Significant SCEPTRE p-values (n=139) are colored in blue and purple, and the top 139 Virtual FACS pairs, as ranked by significance score, are colored in gray-green and purple. The two sets diverged substantially, with only 43% of pairs in shared across sets. b–d These panels are similar to the corresponding panels in Fig. 4. SCEPTRE pairs showed strong enrichment for biological signals associated with enhancer activity on b physical distance, c HIC interaction, and d ChIP-seq metrics relative to other methodsCC 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 34 references

Research works printed but not linked: 1

  1. Zhang L, Janson L. Floodgate : inference for model-free variable importance. arXiv. 2020:1–67.

Cited by

See all 92

Metadata

Authors
Barry, T., Wang, X., Morris, J. A., Roeder, K., & Katsevich, E.
Affiliations
  • Carnegie Mellon University (US)
  • New York Genome Center (US)
  • University of Pennsylvania (US)
Corresponding author
Katsevich, Eugene
Venue
Genome Biology
Published
2021-12
Volume
22
Issue
1
Pages
344-344
ISSN
  • 1465-6906 (ISSN-L)
  • 1474-760X
OpenAlex
W4200438609
Cited by
86
References
34
Citations by year
2026: 28 · 2025: 17 · 2024: 20 · 2023: 14 · 2022: 5 · 2021: 2
FWCI
3.76
Type
journal article
Language
en
Open access
Licence
Locations
  • Genome biology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Institute of Mental Health (R01MH123184, SFARI Grant 575547)
  • National Science Foundation (ACI-1548562, ACI-1445606)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › CRISPR and Genetic Engineering
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene Regulatory Network Analysis
Keywords
single-cell CRISPR screens, scepter, disturbance detection, calibration, sensitivity, regulatory elements, gene expression
MeSH terms
Gene Editing, Chromatin Immunoprecipitation Sequencing, Calibration, Humans, Gene Expression, Genome, Human, Single-Cell Analysis, 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: SCEPTRE improves calibration and sensitivity in single-cell CRISPR screen analysis. https://app.cochat.ai/papers/10.1186/s13059-021-02545-2
The paper
Barry, T., Wang, X., Morris, J. A., Roeder, K., & Katsevich, E. (2021). SCEPTRE improves calibration and sensitivity in single-cell CRISPR screen analysis. Genome Biology. https://doi.org/10.1186/s13059-021-02545-2

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.