Gene and cell line efficiency of CRISPR computed by tensor decomposition in genome-wide CRISPR-Cas9 knockout screens

Abstract

Genome-wide CRISPR-Cas9 knockout screens are often used to experimentally evaluate gene function. However, the efficacy of individual sgRNAs targeting unique genes varies and is difficult to integrate. In this study, tensor decomposition (TD) was used to integrate multiple sgRNAs and sgRNA profiles simultaneously. Thus, TD can discriminate between essential and non-essential genes with the performance comparative to that of Joint analysis of CRISPR/-Cas9 knockout screens (JACKS), a type of SOTA that previously outperformed various other SOTA. In addition, although TD uses simple linear algebra, it can achieve good performance even without control samples, without which JACKS cannot be performed. Moreover, because raw and logarithmic values can achieve similar performances through TD for the largest dataset among the tested datasets, taking logarithmic values as has been done frequently, which is questioned. In conclusion, TD is the first method that can integrate multiple sgRNAs attributed to single a target and sgRNA profiles at the beginning simultaneously and can achieve a performance comparable to that of JACKS.