Biases and Blind-Spots in Genome-Wide CRISPR Knockout Screens

Abstract

It is widely accepted that pooled library CRISPR knockout screens offer greater sensitivity and specificity than prior technologies in detecting genes whose disruption leads to fitness defects, a critical step in identifying candidate cancer targets. However, the assumption that CRISPR 5 screens are saturating has been largely untested. Through integrated analysis of screen data in cancer cell lines generated by the Cancer Dependency Map, we show that a typical CRISPR screen has a ~20% false negative rate, beyond library-specific false negatives previously described. Replicability falls sharply as gene expression decreases, while cancer subtype-specific genes within a tissue show distinct profiles compared to false negatives. Cumulative analyses 10 across tissues suggest only a small number of lineage-specific essential genes and that these genes are highly enriched for transcription factors that define pathways of tissue differentiation.In addition, we show that half of all constitutively-expressed genes are never hits in any CRISPR screen, and that these never-essentials are highly enriched for paralogs. Together these observations strongly suggest that functional buffering masks single knockout phenotypes for a 15 substantial number of genes, describing a major blind spot in CRISPR-based mammalian functional genomics approaches. 65 negatives) and confounding off-target effects (false positives) (Boutros and Ahringer, 2008;Echeverri et al., 2006;Hart et al., 2014).More recently, adaptation of the bacterial CRISPR-Cas9 system to mammalian cells enabled genome-scale approaches to define human essential genes. Studies using this technology 70 revealed that mammalian cells have more essential genes than RNAi screens were able to detect and that, at the same false discovery rate, CRISPR screens generated 3-4 times more essential genes (Hart et al., 2014). Moreover, multiple groups revealed lists of ~2000 highly concordant human essential genes, and comparison of CRISPR technology to orthogonal techniques such as random insertion of gene traps also showed consistent results (Blomen et al., 2015;

Drug mechanism‐of‐action discovery through the integration of pharmacological and CRISPR screens

“…Inhibition of a protein is also intrinsically different than a knockout, as observed for PARP inhibitors whose activity is mediated through formation of cytotoxic PARP‐DNA complexes, whereas PARP knockout has little or no effect on cell fitness (Gill et al , ; Murai & Pommier, ; Antolin et al , ) (Fig EV3H). Similarly, redundancy of gene paralogs when using single‐gene CRISPR knockout may confound comparisons with drugs that target multiple paralogs (preprint: Dede et al , ). Thus, although the reasons may vary for each drug, the absence of an association between drug sensitivity and CRISPR loss‐of‐function effects could warrant further investigation into drug mechanism‐of‐action to understand possible underlying factors, such as low potency, alternative molecular mechanisms, or polypharmacology.…”

CEN‐tools: an integrative platform to identify the contexts of essential genes

“…Our core gene analysis captured almost all genes identified by ADaM (Behan et al , 2019), and identified 195 new core essential genes involved mainly in previously known‐to‐be‐essential cellular processes, but also added an additional process of COP9 signalosome complex to these. The core analysis from CEN‐tools could also capture all but 11 genes from the very recent core gene analysis from Dede et al (preprint: Dede et al, 2020), which used the same data sets as in this study (Appendix Fig S9). As the essentiality measured from genetic screens not only captures genes whose loss causes cell death, but also genes whose loss results in slower proliferation of the cells, the day the experiments were performed will affect the identification of the genes essential for proliferation.…”

Multiplex enCas12a screens detect functional buffering among paralogs otherwise masked in monogenic Cas9 knockout screens

“…The paralog pairs which were both constitutively expressed gene lists were identified and were binned according to different thresholds for percent sequence identity from a range of 10-95%. For each bin, the percentage of constitutively expressed never-essential genes with paralogs and the percentage of common essential genes (defined in [41]) with paralogs were calculated and their distributions were plotted. For downstream analysis, always expressed paralog pair lists were generated for each sequence identity threshold.…”