Skip to main content
Fig. 1 | Genome Biology

Fig. 1

From: ACE: a probabilistic model for characterizing gene-level essentiality in CRISPR screens

Fig. 1

CRISPR-Cas9 Negative Selection Screen and Probabilistic Graphical Model. a The initial pool of Cas9-expressing cells is infected with a lentiviral sgRNA master library. The initial sgRNA abundances are measured by sequencing either the newly infected population of cells, the master library, or both. Cells are allowed to grow, and final sgRNA abundances are obtained by sequencing the surviving cells. b Graphical model illustrating how information is pooled across samples (s) and sgRNAs targeting each gene (gG) to infer each gene’s essentiality, ϕG. From top-left to bottom-right, mg represents the fraction of the master library consisting of sgRNA g; cs represents the number of cells in sample s (estimated by the user); nsg represents the number of cells initially infected with sgRNA g (such that nsgcs); dsg represents the corresponding number of surviving cells after cell growth; and xsg and ysg represent the read counts obtained by sequencing g in the initial and surviving cells, respectively. The numbers of infected cells prior to (nsg) and following (dsg) growth are assumed to be related by the deterministic function dsg=f(nsg;ϕG,εg)=nsg(1−εgϕG) (indicated by factor-graph notation). The efficiency εg is determined by logistic regression and the gene essentiality values ϕG are estimated by maximum likelihood (see Methods). The prior distribution for nsg and the sampling distributions for xsg and ysg are assumed to be Poisson. The scaling factors γs and γs′ accommodate global properties of sequencing depth and cell growth, and are estimated in pre-processing (see the “ Methods” section). Shaded nodes represent variables observed in the data, unfilled nodes represent latent variables, and smaller solid circles represent free parameters

Back to article page