CRISPhieRmix: a hierarchical mixture model for CRISPR pooled screens

Abstract

Pooled CRISPR screens allow researchers to interrogate genetic causes of complex phenotypes at the genome-wide scale and promise higher specificity and sensitivity compared to competing technologies. Unfortunately, two problems exist, particularly for CRISPRi/a screens: variability in guide efficiency and large rare off-target effects. We present a method, CRISPhieRmix, that resolves these issues by using a hierarchical mixture model with a broad-tailed null distribution. We show that CRISPhieRmix allows for more accurate and powerful inferences in large-scale pooled CRISPRi/a screens. We discuss key issues in the analysis and design of screens, particularly the number of guides needed for faithful full discovery.Electronic supplementary materialThe online version of this article (10.1186/s13059-018-1538-6) contains supplementary material, which is available to authorized users.

gscreend: modelling asymmetric count ratios in CRISPR screens to decrease experiment size and improve phenotype detection

“…For the null distribution modelling and hypothesis testing, approaches derived from RNA sequencing and differential gene expression analysis have been used [38,39]. Here, we show that the distribution of the before/after ratios for negative controls is often asymmetric in CRISPR-KO screens; a similar observation has previously been reported for CRISPRi/CRISPRa screens [35]. Such asymmetry means that even in the absence of a fitness effect, a gRNA's relative abundance x is more likely to randomly decrease to, say, x/q (q > 1), rather than increase to xq during the screen.…”

“…To compare the gRNA abundances before and after the proliferation phase, a range of statistical models and computational tools are available [30][31][32][33][34][35][36][37]. Common approaches are to model the joint bivariate null distribution of the normalized counts before and after the proliferation phase, or the null distribution of a univariate summary statistic, the ratio of these counts, hereafter referred to as the "before/after ratio. "…”

“…Since it is common that each gene is targeted by multiple gRNAs, a subsequent step of the analysis consists of aggregating gRNA-level evidence to the gene level. This can be achieved for example using Bayesian hierarchical modelling [31,35,36] or robust rank aggregation [30] (see also Additional file 1: Table S1 for a short summary of the methods compared in this work).…”

A benchmark of algorithms for the analysis of pooled CRISPR screens

“…Daley et al developed CRISPhieRmix [28] to address issues in the analysis of CRISPR activation and inactivation screens, specifically the issue of variable guide efficiency that are ubiquitous in these screens. CRIS-PhieRmix takes as input the log fold changes of the sgRNA from the initial condition to the final condition, typically estimated by standard count software such as DESeq2 [36] or edgeR [39], and then fits a hierarchical mixture distribution, assuming that the guides for hit genes can follow a mixture distribution.…”

“…Despite great utility, challenges still remain in CRISPR screens. Variability in guide RNA efficiency can complicate the analysis [28]. Varying gene effect sizes can result in a bias towards finding only genes with large effects [29].…”