Compact and highly active next-generation libraries for CRISPR-mediated gene repression and activation

  1. Max A Horlbeck
  2. Luke A Gilbert
  3. Jacqueline E Villalta
  4. Britt Adamson
  5. Ryan A Pak
  6. Yuwen Chen
  7. Alexander P Fields
  8. Chong Yon Park
  9. Jacob E Corn
  10. Martin Kampmann
  11. Jonathan S Weissman  Is a corresponding author
  1. University of California, San Francisco, United States
  2. University of California, Berkeley, United States
  3. University of California, San Francisco, United states
4 figures and 10 additional files

Figures

Figure 1 with 2 supplements
A machine learning approach for identifying highly active sgRNAs for CRISPRi.

(A) Schematic of machine learning strategy and datasets. 808 features were calculated for each sgRNA, linearized as indicated, and z-standardized. A linear regression model was then generated using …

https://doi.org/10.7554/eLife.19760.002
Figure 1—figure supplement 1
Relationship between CRISPRi activity and sgRNA position relative to the TSS as predicted by SVR.

An SVR model with radial basis function kernel was trained on an 80% division of the CRISPRi activity score dataset, using the position of the sgRNA relative to the upstream end of the primary …

https://doi.org/10.7554/eLife.19760.003
Figure 1—figure supplement 2
Individual sgRNA feature contributions to predicted CRISPRi activity.

Linear regression coefficients for each model trained according to the 80%/20% divisions displayed in Figure 1 are displayed, with bars indicating the mean of the five divisions and error bars …

https://doi.org/10.7554/eLife.19760.004
Figure 2 with 2 supplements
A machine learning approach for identifying highly active sgRNAs for CRISPRa.

(A) Schematic of CRISPRa datasets. CRISPRa activity scores were generated from screen data and subjected to 3-fold cross-validation due to the smaller sample size. Ricin tiling data was limited to …

https://doi.org/10.7554/eLife.19760.005
Figure 2—figure supplement 1
Relationship between CRISPRa activity and sgRNA position relative to the TSS as predicted by SVR.

An SVR model was trained on a 67% division of the CRISPRa dataset using sgRNA position relative to the downstream end of the primary FANTOM TSS for each gene. Analysis was conducted as described for …

https://doi.org/10.7554/eLife.19760.006
Figure 2—figure supplement 2
Individual sgRNA feature contributions to predicted CRISPRa activity.

Linear regression coefficients for each model trained according to the 67%/33% divisions displayed in Figure 2 are displayed, with bars indicating the mean of the three divisions and error bars …

https://doi.org/10.7554/eLife.19760.007
Figure 3 with 2 supplements
hCRISPRi-v2 outperforms CRISPRi v1 in screens for essential genes in K562.

(A) Distribution and predicted scores for sgRNAs targeting essential genes. (Top) Predicted activity scores for sgRNAs from Evers et al., 2016 or hCRISPRi-v2 targeting essential genes as defined by …

https://doi.org/10.7554/eLife.19760.008
Figure 3—figure supplement 1
sgRNA phenotypes from CRISPRi v1 and hCRISPRi-v2 growth screens.

(A) hCRISPRi-v2 sgRNA phenotypes correlate between screen replicates. sgRNA γ scores were calculated by computing log2 enrichments of read counts between screen start and endpoint samples and …

https://doi.org/10.7554/eLife.19760.009
Figure 3—figure supplement 2
Precision-recall analysis of second-generation CRISPR nuclease essential gene screens.

Analysis was conducted as in Figure 3D. CRISPRi, Wang et al., and Doench et al. datasets were ranked according to the average log2 fold-change of all sgRNAs targeting a given gene. For Hart et al., …

https://doi.org/10.7554/eLife.19760.010
Figure 4 with 1 supplement
hCRISPRa-v2 outperforms CRISPRa v1 in screens for genes that modify growth rates upon overexpression.

(A) Volcano plots of gene phenotypes and p-values for growth screens performed with CRISPRa v1 (Gilbert et al., 2014) and hCRISPRa-v2, presented as in Figure 3C. (B) Cumulative distributions of …

https://doi.org/10.7554/eLife.19760.011
Figure 4—figure supplement 1
sgRNA phenotypes and gene category enrichment scores from CRISPRa v1 and hCRISPRa-v2 growth screens.

(A) hCRISPRa-v2 sgRNA phenotypes correlate between screen replicates, presented as in Figure 3—figure supplement 1A. Phenotypes for non-targeting sgRNAs and sgRNAs targeting the Y chromosome …

https://doi.org/10.7554/eLife.19760.012

Additional files

Supplementary file 1

CRISPRi and CRISPRa activity score datasets.

https://doi.org/10.7554/eLife.19760.013
Supplementary file 2

TSS annotations for hg19 and mm10 genomes.

https://doi.org/10.7554/eLife.19760.014
Supplementary file 3

Library composition of hCRISPRi-v2 and hCRISPRi-v2.1.

https://doi.org/10.7554/eLife.19760.015
Supplementary file 4

Library composition of mCRISPRi-v2.

https://doi.org/10.7554/eLife.19760.016
Supplementary file 5

Library composition of hCRISPRa-v2.

https://doi.org/10.7554/eLife.19760.017
Supplementary file 6

Library composition of mCRISPRa-v2.

https://doi.org/10.7554/eLife.19760.018
Supplementary file 7

sgRNA read counts and growth phenotypes for hCRISPRi-v2 screens performed in K562.

https://doi.org/10.7554/eLife.19760.019
Supplementary file 8

Gene growth phenotypes and p-values for hCRISPRi-v2 screens performed in K562.

https://doi.org/10.7554/eLife.19760.020
Supplementary file 9

sgRNA read counts and growth phenotypes for hCRISPRa-v2 screens performed in K562.

https://doi.org/10.7554/eLife.19760.021
Supplementary file 10

Gene growth phenotypes and p-values for hCRISPRa-v2 screens performed in K562.

https://doi.org/10.7554/eLife.19760.022

Download links