Variation in guide RNA library representation results in gene effect score bias in genome-wide CRISPR screens

Genome wide CRISPR-based perturbation screens are powerful discovery tools enabling the identification of novel gene dependencies through either gain or loss of function. While genome wide guide RNA (gRNA) libraries have advantages when using enAsCas12a, such as multiplex single gRNAs per gene, they may be subject to similar confounding factors that can affect the interpretation of large genome-wide datasets. Here, we examine the impact of these variables in over twenty enAsCas12a multiple gRNA based perturbation screens performed using Humagne C, Humagne D and Inzolia libraries in human cells, as well as external datasets containing Cas9-based CRISPR library screens, including from DepMap. We demonstrate that the choice of CRISPR library is often the most significant factor that influences genetic perturbation results, outweighing other variables such as either target cell lines or culture media conditions. A potential contributor to this effect is gRNA representation within a given CRISPR library, where lower gRNA representation can lead to variable and more pronounced gene effect scores using either log fold change or Chronos analysis. These effects may be mitigated by using either multiple gRNA constructs per gene, by optimisation of CRISPR library production processes or by targeting with multiple independent gRNA libraries. Importantly, we also propose strategies for addressing gRNA representation bias during CRISPR screen hit prioritisation. CRISPR library gRNA representation dependent bias remains a major challenge in the interpretation of gene essentiality in perturbation screens.