The scope of the L1000 dataset provides an unprecedented opportunity to examine the biological effects of shRNAs, in particular, their off-target effects. We analyzed 13,187 shRNAs targeting 3,799 genes across 9 cell lines, and compared each pair of shRNA-induced L1000 profiles, comparing similarity between shRNAs targeting the same gene (“shared gene”) and shRNAs targeting different genes but sharing the 2-8 nucleotide seed sequence known to contribute to off-target effects (“shared seed”) (Jackson et al., 2003). Figure 3A shows that shared gene similarity is only slightly greater than random. In contrast, shared seed pairs were dramatically more similar compared to the null distribution, indicating that the magnitude of off-target effects of shRNAs substantially exceeds the magnitude of their on-target effect. We reasoned that while on-target gene expression effects of different shRNAs targeting the same gene should be the same, their off-target effects should not. We therefore developed an algorithm to produce a Consensus Gene Signature (CGS) that reflects the consistent (and therefore on-target) gene expression effects of shRNAs and used the CGS output for all analyses that follow. The CGS method and its validation are described in detail elsewhere (Smith et al., 2017) .
