CoChat
Sign Up
Cited by 278CRISPR and Genetic EngineeringSingle-cell and spatial transcriptomicsVirus-based gene therapy research

On the design of CRISPR-based single-cell molecular screens

Open access
Hill, Andrew J, McFaline-Figueroa, José L, Starita, Lea M, Gasperini, Molly J, Matreyek, Kenneth A, Packer, Jonathan, Jackson, Dana, Shendure, Jay +1 more·2018·Nature Methods
DOI 10.1038/nmeth.4604PMC PMC5882576

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Several groups recently coupled CRISPR perturbations and single-cell RNA-seq for pooled genetic screens. We demonstrate that vector designs of these studies are susceptible to ∼50% swapping of guide RNA-barcode associations because of lentiviral template switching. We optimized a published alternative, CROP-seq, in which the guide RNA also serves as the barcode, and here confirm that this strategy performs robustly and doubled the rate at which guides are assigned to cells to 94%.

AI snapshot

Written from the paper’s full text.

TL;DR

This paper shows that common CRISPR single-cell screen vectors that pair a guide RNA with a distant barcode can suffer roughly 50% guide–barcode reassortment from lentiviral template switching, which sharply reduces differential-expression power. The authors improve CROP-seq by targeted amplification of the guide RNA, raising guide-to-cell assignment to 94% while avoiding this swapping problem.

Key findings

  • In a two-plasmid BFP/GFP assay with 2.4 kb separation, barcode swapping reached the theoretical maximum of 50% under pooled lentiviral production.
  • In silico label permutation at 50% swapping reduced detected differentially expressed genes 4.8-fold in data from Adamson et al. and 2.9-fold in the authors’ own CROP-seq tumor-suppressor screen; significant targets fell to 4/13 in mock and 3/14 in doxorubicin conditions.
  • Adding targeted guide-RNA amplification to CROP-seq increased guide assignment from 42–47% with scRNA-seq alone to 94%.
  • In the improved CROP-seq doxorubicin experiment, a TP53-enriched cluster contained 262 cells, of which 90.5% had TP53-targeting guides, 7.6% had guides to other genes, 0% had non-targeting controls, and 1.9% were unassigned; CDKN1A and TP53I3 were lower in that cluster.

Methods

  • The authors used MCF10A breast epithelial cells with inducible or constitutive Cas9, arrayed and pooled lentiviral production, and doxorubicin exposure to compare vector designs and screen tumor suppressors.
  • They performed 10x Chromium scRNA-seq with separate barcode/guide enrichment PCR, then assigned guides or barcodes to cells and analyzed expression with Monocle2, t-SNE, differential expression, and gene set enrichment.
  • They simulated template switching by permuting target labels at rates from 0–100% and counting differentially expressed genes at 5% FDR, including on published Perturb-seq data.

Limitations

  • The paper states that CROP-seq is limited by placing the sgRNA in the lentiviral LTR, which may not tolerate larger intervening sequences such as dual sgRNA designs.
  • The authors report that their pHAGE-scKO design had poor CRISPRi efficacy and low editing rates (29% vs. 88% for a control), hypothesizing interference from overlapping Pol II and Pol III transcripts.
  • The paper notes that computational filtering of inconsistent cells can introduce bias and discard data, while arrayed lentiviral production limits scalability; it also states that positive conclusions from prior pooled studies are not expected to be incorrect, though power may drop as target numbers grow.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/nmeth.4604

Cites

31 of 31 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 30 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    When target labels in the CROP-seq tumor suppressor screen were shuffled at a 50% rate, DEG counts across targets fell 2.9-fold.

    “To further assess the impact of template switching on sensitivity, we permuted target labels within our own CROP-seq tumor suppressor screen, observing a 2.9-fold reduction in the number of DEGs across targets at a swap rate of 50%.”
    Quote in the paperSubstantial claim
  • Findingown data

    The TP53-targeted cluster showed much lower expression of the p53 targets CDKN1A and TP53I3.

    “Expression levels of the p53 targets CDKN1A and TP53I314,15 were markedly lower in the TP53-targeted cluster (Figure 2c)”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [14] el-Deiry, W.S. et al. Cell 75, 817–825 (1993). source

      Not checked in the free run.

    • cites [15] Contente, A., Dittmer, A., Koch, M.C., Roth, J. & Dobbelstein, M. Nat. Genet. 30, 315–320 (2002). source

      Not checked in the free run.

  • Findingown data

    The pooled experiment yielded far fewer differentially expressed genes than the arrayed experiment.

    “Furthermore, markedly fewer genes were differentially expressed in the pooled than in the arrayed experiment (Supplementary Figure 3c).”
    Quote in the paperSubstantial claim
  • Findingown data

    In the arrayed experiment, nearly all cells in that TP53-loss-like cluster carried sgRNAs against TP53 (99.4%).

    “In the arrayed experiment, this cluster was almost entirely composed of cells with sgRNAs targeting TP53 (99.4%).”
    Quote in the paperSubstantial claim
  • Findingown data

    Applying a 50% swap rate to Adamson et al.'s Perturb-seq data reduced the count of differentially expressed genes by 4.8-fold.

    “With 50% swapping, we observe a 4.8-fold decrease in the number of DEG (Figure 1f).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [4] Adamson, B. et al. Cell 167, 1867–1882.e21 (2016). source

      Not checked in the free run.

  • Classified as probably not a substantial claim. 15
  • May combine two claims. 19
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    For the initial arrayed and pooled doxorubicin-treated pLGB-scKO samples, cellranger aggregate normalization produced per-cell averages of 37,732 reads, median gene counts of 2,263, and median UMI counts of 8,279.

    “This yields an average of 37,732 reads per cell, 2263 median genes per cell, and a median of 8279 UMIs per cell.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data

    When lentivirus was produced individually for each gene, a large fraction of TP53-targeted cells showed expression changes resembling a defective cell-cycle checkpoint after DNA damage, with reduced CDKN1A and TP53I3.

    “With arrayed lentiviral production, a substantial proportion of cells in which TP53 was targeted had a gene expression signature consistent with failure to activate a cell cycle checkpoint response after DNA damage (e.g., lower expression of CDKN1A and TP53I3; Supplementary Figure 3a).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    For the CROP-seq mock sample, sequencing yielded 120,797 raw reads per cell on average across 6,598 cells; median detected genes were 4,619 and median UMIs were 22,495 per cell.

    “Our CROP-seq mock sample was sequenced to an average depth of 120,797 raw reads per cell in 6598 cells. A median of 4619 genes per cell were detected and a median UMI count of 22,495 per cell.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    At the sequencing depth of the CROP-seq samples, the average duplication rate was approximately 78%.

    “At this depth the average duplication rate is approximately 78%.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    With 50% swapping, significant targets dropped to 4 of 13 (TP53, STK11, CHEK1, NCOR1) in mock and 3 of 14 (TP53, RB1, ARID1B) after doxorubicin.

    “The number of significant targets was also reduced, to just 4/13 (TP53, STK11, CHEK1 and NCOR1) and 3/14 (TP53, RB1, and ARID1B) in the mock and doxorubicin conditions, respectively.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    That cluster contained 262 cells: 90.5% had TP53 guides, 7.6% had guides for other genes, none had NTC guides, and 1.9% were unassigned.

    “Specifically, the 262 cells in this cluster include 90.5% with TP53-targeting guides, 7.6% with guides targeting other genes, 0% with NTC guides, and 1.9% unassigned cells.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Among the other 5,617 cells, 3.2% carried TP53 guides, 84.2% carried guides for other genes, 7.5% carried NTC guides, and 5.2% were unassigned.

    “In contrast, the remaining 5,617 cells include 3.2% with TP53-targeting guides (presumably cells in which LoF editing failed to occur), 84.2% with guides targeting other genes, 7.5% with NTC guides, and 5.2% unassigned cells.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    For the CROP-seq doxorubicin-treated sample, sequencing yielded 123,445 raw reads per cell on average across 6,283 cells; median detected genes were 3,500 and median UMIs were 15,324 per cell.

    “Our CROP-seq doxorubicin treated sample was sequenced to an average depth of 123,445 raw reads per cell in 6283 cells. A median of 3500 genes per cell were detected and we observed a median UMI count of 15,324 per cell.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The Dixit-based removal of likely chimeric sequences affects the raw data somewhat, though the advantages are modest.

    “While this has some impact on the raw data, we find the benefits to be modest.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Dixit24

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    The improved CROP-seq method, using the guide RNA itself as the barcode, works reliably and raises guide-to-cell assignment to 94%, about twice the previous rate.

    “We optimize a published alternative, CROP-seq, in which the guide RNA also serves as the barcode, confirming this strategy performs robustly and doubling the rate at which guides are assigned to cells to 94%.”
    • Also stated in: § Abstract · plain
      “While sgRNA(s) would be identified at a rate of 42-47% from scRNA-seq data alone, this rate was increased to 94% with targeted amplification (Figure 2a).”
    • Also stated in: § Abstract · plain
      “By coupling targeted sgRNA amplification and CROP-seq, we doubled the proportion of cells in which guides are assigned to 94%.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In a CRISPRi test against an mCherry promoter, pKHH030 and CROP-seq inhibited efficiently, but pHAGE-scKO worked poorly.

    “Whereas pKHH030 and CROP-seq exhibited efficient inhibition, pHAGE-scKO had poor efficacy (Figure 2d).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    t-SNE of mock and doxorubicin-treated cells showed that several tumor suppressors had cluster distributions differing significantly from NTCs (FDR 5%); more targets changed after doxorubicin (13 in mock, 14 in doxorubicin).

    “Upon tSNE analysis of both mock and doxorubicin treated cells (Supplementary Figure 8a-b), we find several tumor suppressors whose distribution across clusters is significantly different compared to NTCs (FDR 5%), with more changes after doxorubicin exposure (13 and 14 targets with significant changes in the mock and doxorubicin conditions, respectively; Supplementary Fig. 8c–f).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The sorting instrument was adjusted to correct for spectral overlap between EBFP and EGFP.

    “The mixed cells along with others were sorted for blue+ or green+ populations using a FACS Aria II (Becton Dickinson) that had been compensated for the overlap between the EBFP and EGFP emission spectra.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Adamson et al. eliminated this problem by producing lentivirus individually, while other studies used pooled production in at least some experiments.

    “It was altogether avoided by Adamson et al.4 through arrayed lentiviral production, but pooled lentiviral production was performed in some or all experiments of the other reports3,5,6.”
    Quote in the paperSubstantial claim0 of 4 sources checked
    • cites [4] Adamson, B. et al. Cell 167, 1867–1882.e21 (2016). source

      Not checked in the free run.

    • cites [3] Xie, S., Duan, J., Li, B., Zhou, P. & Hon, G.C. Mol. Cell 66, 285–299.e5 (2017). source

      Not checked in the free run.

    • cites [5] Dixit, A. et al. Cell 167, 1853–1866.e17 (2016). source

      Not checked in the free run.

    • cites [6] Jaitin, D.A. et al. Cell 167, 1883–1896.e15 (2016). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    Swapping weakens detection because each target effectively has half as many informative cells and because mismatched labels add noise.

    “This loss in power results from an effective twofold reduction in number of useful cells per target, coupled with noise from swapped associations.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Adamson et al. and Dixit et al. had already pointed out this template-switching risk.

    “This risk was noted by Adamson et al.4 and Dixit et al.5.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [4] Adamson, B. et al. Cell 167, 1867–1882.e21 (2016). source

      Not checked in the free run.

    • cites [5] Dixit, A. et al. Cell 167, 1853–1866.e17 (2016). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Barcodes in individual clones were verified by Sanger sequencing.

    “Single clones were prepared and the barcode identified by Sanger sequencing.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The pHAGE-scKO design should prevent sgRNA copying between LTRs while still allowing the guide sequence to appear in overlapping Pol II and Pol III transcripts.

    “In this design (pHAGE-scKO), copying of the sgRNA between LTRs would not occur, but the guide sequence would still contribute to overlapping Pol II and Pol III transcripts (Supplementary Figure 11).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Individual clones were validated by Sanger sequencing.

    “Individual clones were validated via Sanger sequencing.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    In a two-plasmid mixture, only half of chimeric products are detectable, because many virions are homozygous and their chimeric products match the original construct.

    “Note that, unlike a library of plasmids, in a mix of two plasmids, only half of all chimeric products will be detectable as many virions will be homozygous (i.e., contain the same construct, and thus chimeric products are identical to the original).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    The modified CROP-seq method matches the arrayed format's sensitivity and very low sgRNA swapping while keeping the scalability of pooled cloning and lentiviral production.

    “Thus, our improved CROP-seq protocol achieves the power and negligible sgRNA swap rate of the arrayed format without sacrificing the scalability of a pooled cloning and lentiviral production workflow.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Equal numbers of cells receiving EGFP and EBFP virus were combined to measure contamination caused by sorting error.

    “An equal number of cells transduced with EGFP and EBFP virus were mixed to determine the rate of contamination resulting from FACS error.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    CROP-seq avoids sgRNA-barcode swapping, but its sgRNA sits in the lentiviral LTR, and longer inserts such as dual sgRNA cassettes could make that LTR non-functional.

    “Although CROP-seq is not subject to sgRNA-barcode swapping, it is limited by its placement of the sgRNA in the lentiviral LTR, as larger intervening sequences such as dual sgRNA designs16 might render the LTR non-functional7.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [16] Gasperini, M. et al. Am. J. Hum. Genet. 101, 192–205 (2017). source

      Not checked in the free run.

    • cites [7] Datlinger, P. et al. Nat. Methods 14, 297–301 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The pHAGE_dsRed_IRES_zsGreen_WPRE_MCS vector clones were chosen using Sanger sequencing and fluorescent protein expression after transfection and packaging.

    “Clones of pHAGE_dsRed_IRES_zsGreen_WPRE_MCS were chosen by Sanger sequencing and expression of the fluorescent proteins after transfection and lentiviral packaging.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Transduced cells were maintained for four weeks so that leftover plasmid would not contaminate the results.

    “The cells were cultured for four weeks to avoid residual plasmid contamination following transduction.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    CROP-seq recovers sgRNAs from scRNA-seq data at only about 40-60% sensitivity, so roughly half of the single-cell transcriptomes are lost.

    “A limitation of CROP-seq is that sgRNAs are recovered from scRNA-seq data with limited sensitivity (~40-60%)7, such that half the single cell transcriptomes are discarded.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [7] Datlinger, P. et al. Nat. Methods 14, 297–301 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The swap-rate analysis was rerun with the GFP+ fraction left free rather than set to the FACS value, to verify concordant results.

    “This analysis was also performed without fixing the fraction of GFP+ cells to the value measured by FACS to ensure results were concordant (Supplementary Figure 6).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Sack et al. measured this phenomenon for separations as long as 720 bp in vectors intended for bulk selection.

    “Although Sacket al. recently quantified this phenomenon at distances up to 720 bp in vectors designed for bulk selection screens12”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [12] Sack, L.M., Davoli, T., Xu, Q., Li, M.Z. & Elledge, S.J. G3 (Bethesda) 6, 2781–2790 (2016). source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductreasoning

    The editing-rate experiment did not use UMIs, so the measured editing rate may be overestimated.

    “We did not use UMIs in this experiment and thus it may overestimate editing rate.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Functional editing-rate estimates would be confounded if one accounted for target-gene loss-of-function edits whose cellular phenotypic effects are incomplete.

    “Such estimates would be confounded if accounting for the possibility of edits that cause LoF for the target gene but have incomplete penetrance on the cellular phenotype.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning

    How template switching affects longer separations, including the >2.5 kb gaps between sgRNAs and barcodes in pLGB-scKO, CRISP-seq, Perturb-seq, and Mosaic-seq, and its consequences for scRNA-seq designs, had not been studied.

    “the implications of template switching at longer distances (e.g. the 2.5 kb+ separation between sgRNAs and barcodes in the pLGB-scKO, CRISP-seq, Perturb-seq, and Mosaic-seq vectors3–6), as well as for scRNA-seq study designs specifically, remain unexplored.”
    Quote in the paperSubstantial claim0 of 4 sources checked
    • cites [3] Xie, S., Duan, J., Li, B., Zhou, P. & Hon, G.C. Mol. Cell 66, 285–299.e5 (2017). source

      Not checked in the free run.

    • cites [4] Adamson, B. et al. Cell 167, 1867–1882.e21 (2016). source

      Not checked in the free run.

    • cites [5] Dixit, A. et al. Cell 167, 1853–1866.e17 (2016). source

      Not checked in the free run.

    • cites [6] Jaitin, D.A. et al. Cell 167, 1883–1896.e15 (2016). source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Unlike the other approaches, CROP-seq does not pair sgRNAs with separate barcodes; the sgRNA is itself the barcode within an overlapping Pol II transcript.

    “CROP-seq7 differs from the other methods3–6 in not relying on pairing of sgRNAs and barcodes. Instead, the sgRNA itself serves as a barcode, as part of an overlapping Pol II transcript.”
    Quote in the paperSubstantial claim0 of 5 sources checked
    • cites [7] Datlinger, P. et al. Nat. Methods 14, 297–301 (2017). source

      Not checked in the free run.

    • cites [3] Xie, S., Duan, J., Li, B., Zhou, P. & Hon, G.C. Mol. Cell 66, 285–299.e5 (2017). source

      Not checked in the free run.

    • cites [4] Adamson, B. et al. Cell 167, 1867–1882.e21 (2016). source

      Not checked in the free run.

    • cites [5] Dixit, A. et al. Cell 167, 1853–1866.e17 (2016). source

      Not checked in the free run.

    • cites [6] Jaitin, D.A. et al. Cell 167, 1883–1896.e15 (2016). source

      Not checked in the free run.

32 of 32

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?

  • SenCID can be used in high-throughput single-cell CRISPR screens to identify genome-wide senescence trigger and suppressor genes.

    “Next, we investigate whether SenCID can be used in high-throughput single-cell CRISPR screens, such as CROP-seq117,118 or Perturb-seq119,120 to identify genome-wide senescence trigger and suppressor genes.”

    From Single-cell senescence identification reveals senescence heterogeneity, trajectory, and modulators (2024) · cites this paper as [118]

  • Current single-cell CRISPR screening implementations have limitations including poor scalability, dependence on specialized vector systems, and high cost.

    “current implementations face technical and practical limitations that unnecessarily restrict their use. Here, we present advances that address these limitations, specifically poor scalability, dependence on specialized vector systems and high cost9–12”

    From Combinatorial single-cell CRISPR screens by direct guide RNA capture and targeted sequencing (2020) · cites this paper as [9]

  • During lentiviral delivery, recombination can separate indexes from the sgRNAs they were assigned to in indexed sgRNA libraries.

    “However, recombination of indexed sgRNA libraries during lentiviral delivery can uncouple indexes from their assigned sgRNAs9–12.”

    From Combinatorial single-cell CRISPR screens by direct guide RNA capture and targeted sequencing (2020) · cites this paper as [9]

  • Direct capture Perturb-seq can identify cells bearing undesired sgRNA pairs generated from intermolecular lentiviral recombination and computationally exclude them from downstream analysis.

    “Lastly, with direct capture Perturb-seq, we can identify cells bearing undesired sgRNA pairs (generated from intermolecular lentiviral recombination between programmed pairs9–12) and computationally exclude them from downstream analysis.”

    From Combinatorial single-cell CRISPR screens by direct guide RNA capture and targeted sequencing (2020) · cites this paper as [9]

  • TAP-seq recovered gRNA identity at 95% efficiency, versus 39% for Perturb-seq and 89% for Perturb-seq with targeted gRNA amplification.

    “TAP-seq captured gRNA identity with an efficiency of 95%, as compared to 39% in the Perturb-seq setting or 89% in Perturb-seq with additional targeted gRNA amplification20”

    From Targeted Perturb-seq enables genome-scale genetic screens in single cells (2020) · cites this paper as [20]

Self-check

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

Each contradiction between two passages listed here was confirmed by an independent check.

  • Wild-type fraction is defined using non-wild-type alleles

    Minor

    The method names the quantity a wild-type genotype fraction but defines it as the fraction of non-wild-type alleles. Those are complementary quantities, not the same fraction.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Dixit, A. Preprint at https://www.biorxiv.org/content/early/2016/12/12/093237 (2016).

Cited by

Sorted by citations

Related

Authors
Hill, A. J., McFaline-Figueroa, J. L., Starita, L. M., Gasperini, M. J., Matreyek, K. A., Packer, J., Jackson, D., Shendure, J., & Trapnell, C.
Affiliations
  • University of Washington (US)
  • Howard Hughes Medical Institute (US)
Corresponding author
  • Shendure, Jay
  • Trapnell, Cole
ORCID
Venue
Nature Methods
Publisher
Springer Science and Business Media LLC
Published
2018-02-19
Volume
15
Issue
4
Pages
271-274
ISSN
  • 1548-7091 (ISSN-L)
  • 1548-7105
DOI
10.1038/nmeth.4604
PMC
PMC5882576
OpenAlex
W2950261377
Cited by
262
References
31
Citations by year
2026: 38 · 2025: 33 · 2024: 45 · 2023: 32 · 2022: 27 · 2021: 24 · 2020: 28 · 2019: 22 · 2018: 13
FWCI
8.00
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Methods (publishedVersion)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nd)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (acceptedVersion)
  • PubMed Central (acceptedVersion)
Funders
  • National Science Foundation
  • W. M. Keck Foundation
  • National Institutes of Health
  • National Heart, Lung, and Blood Institute (T32HL007828)
  • National Human Genome Research Institute (DP1HG007811, UM1HG009408, T32HG000035)
  • Eunice Kennedy Shriver National Institute of Child Health and Human Development (DP2HD088158)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › CRISPR and Genetic Engineering
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Virus-based gene therapy research
Keywords
CRISPR screen, single-cell RNA sequencing, vector design
MeSH terms
RNA, Guide, CRISPR-Cas Systems, Lentivirus, Sequence Analysis, RNA, 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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-04.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-04.

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