Integrated cross-study datasets of genetic dependencies in cancer

Abstract

CRISPR-Cas9 viability screens are increasingly performed at a genome-wide scale across large panels of cell lines to identify new therapeutic targets for precision cancer therapy. Integrating the datasets resulting from these studies is necessary to adequately represent the heterogeneity of human cancers and to assemble a comprehensive map of cancer genetic vulnerabilities. Here, we integrated the two largest public independent CRISPR-Cas9 screens performed to date (at the Broad and Sanger institutes) by assessing, comparing, and selecting methods for correcting biases due to heterogeneous single-guide RNA efficiency, gene-independent responses to CRISPR-Cas9 targeting originated from copy number alterations, and experimental batch effects. Our integrated datasets recapitulate findings from the individual datasets, provide greater statistical power to cancer- and subtype-specific analyses, unveil additional biomarkers of gene dependency, and improve the detection of common essential genes. We provide the largest integrated resources of CRISPR-Cas9 screens to date and the basis for harmonizing existing and future functional genetics datasets.

CoRe: A robustly benchmarked R package for identifying core-fitness genes in genome-wide pooled CRISPR-Cas9 screens

“…We compared the sets of CFGs and CEGs predicted by CoRe when applied to the largest integrative dataset of cancer dependency assembled to date [20] with state-of-the-art sets of core-fitness genes derived from recent functional genetic screening datasets [10,12,18,19], as well as with the output of a logistic-regression based method, part of the recent CEN-tools software proposed in [18], applied to the same dataset [20].…”

“…curated in [27] and [20] from MsigDB [28]. As negative controls we assembled a set of genes never expressed (fragments per kilobase of transcript per million mapped reads (FPKM) < 0.1) in more than 1,000 human cancer cell lines (from the Cell Model Passports [22]), or whose fitness signal across hundreds of cell lines has a t-skewed normal distribution (according to the normLRT score introduced and applied to an independent shRNA-based cancer dependency dataset in [29]) and it is statistically associated with a genomic marker [20]. Excluding genes included in at least one of the training sets yielded a final set of 408 positive controls and 7,767 negative controls ( Supplementary Table 2 ).…”

“…A gene is unexpressed if its FPKM scores (dataset available at https://cellmodelpassports.sanger.ac.uk/downloads, version: rnaseq_20191101) are constitutively less than 0.1 across all the cell lines. The biomarker/dependency associations were derived from [20]. The metrics derived from the baseline predictor were used to assess each CF set.…”

“…We ran BAGEL v115 on the Sanger release 1 cancer dependency dataset (downloaded from: https://score.depmap.sanger.ac.uk/downloads) processed with CRISPRcleanR [27] (shown in [20] to better preserve context-specific essentialities than CERES). As a positive training gene set we used each of the sets among state-of-the-art sets, or CFG sets derived from the supervised methods, in turn, whereas as a negative training gene set we used the curated BAGEL never-essential genes.…”

“…We present CoRe : an R package implementing recently proposed as well as novel versions of algorithms for the identification of CFGs from a joint analysis of multiple genome-wide pooled CRISPR-Cas9 knock-out screens. Furthermore, we present results from a comparison of CoRe’s output (when applied to the largest integrative cancer dependency dataset generated to date [20]) against widely used [10,19], or more recent [18] sets of CFGs obtained via an alternative approach (also tested on the same recent cancer dependency dataset). We report an increased coverage of prior known human essential genes, new potential core-fitness genes, and lower false positive rates for CoRe’s methods with respect to other state-of-the-art core-fitness sets and available methods.…”