Skip to main content
Fig. 1 | Genome Biology

Fig. 1

From: CRISPhieRmix: a hierarchical mixture model for CRISPR pooled screens

Fig. 1

a A hypothetical situation. Gene A has a single sgRNA with a large effect, while the rest of the sgRNAs have little effect. Gene B has very few sgRNAs with large effects, but the distribution looks shifted from the negative control. Gene C appears to follow a clear mixture distribution, with some percentage of the sgRNAs having a large effect but the majority looking like the negative control distribution. b The hierarchical structure of CRISPhieRmix. Genes can either have no association with the phenotype, in which case they are not interesting, or they are interesting and deserve further investigation. Without prior knowledge, the genes can be considered as arising from a mixture distribution. For genes that are interesting, we model sgRNAs as a mixture distribution as well, with some sgRNAs having no effect and following the null distribution and the others having some effect and following the alternative distribution. c Overview of the CRISPhieRmix algorithm. First, a broad-tailed null distribution is fit using negative control sgRNAs. Then, a mixture distribution is fit on all sgRNAs, ignoring gene identities. Finally, for each gene, CRISPhieRmix uses the mixture distribution to calculate the local false discovery rate, equal to the posterior probability that the gene is null. Of the three genes in a, we estimate that gene A is highly likely to be null, gene B is highly likely to be a true hit, and gene C is probably non-null

Back to article page