Our analysis of 18,493 shRNA profiles showed that the off-target effects of shRNAs far exceed their on-target effects, consistent with recent reports (Tsherniak et al., 2017). However, the generation of a Consensus Gene Signature (CGS) that identifies gene expression changes common to multiple shRNAs targeting the same gene substantially improved the ability to discover on-target connections by minimizing off-target effects. Nevertheless, the CGS procedure is imperfect, and some off-target effects likely remain. Preliminary studies of CRISPR/Cas9-mediated gene knock-out suggest that genome editing approaches may recover some genetic connections to small molecules that were missed by RNA interference-based perturbation. Two caveats bear mentioning. First, we and others have shown that CRISPR/Cas9-based genome editing results in non-specific toxicity directly proportional to the number of cuts to the genome (Aguirre et al., 2016). The extent to which such non-specific effects can be computationally corrected in the context of CMap analysis remains to be determined. This is particularly relevant when performing genetic perturbations in cancer cell lines that often harbor copy number alterations. Second, it remains to be determined whether complete gene knock-out (via CRISPR) or partial knock-down (via shRNA) better phenocopies the effect of a small molecule.
