Fig. 5
From: INFIMA leverages multi-omics model organism data to identify effector genes of human GWAS variants
INFIMA model overview. a Input data for the INFIMA model. INFIMA leverages summaries of model organism multi-omics data to model the relationship between allelic expression patterns (Yg) of DO-eQTL genes (i) and founder expression patterns (Bg) under a null model of no causal SNPs (Vg=0); (ii) and founder genotype expression patterns Rg=Eg⊤Zg, where Eg represents genotype effects of candidate SNPs on founder expression and Zg encodes the causal SNP for gene g, under an alternative model with causal SNPs (Vg=1), across all the genes indexed by g. b Plate representation of the INFIMA model summarizing data and the parameters. Blank circles: latent variables and parameters to be inferred; filled circles: observed variables. c INFIMA infers posterior probabilities of association for fine-mapping across all the candidate local-ATAC-MVs. d–g An example input of the INFIMA model. d An overview of a W= 1 Mb window around a DO-eQTL marker (centered dashed line) associated with Gene1. Two out of five candidate local-ATAC-MVs (red short lines) are decorated with comparative footprint effects (orange triangles). e Example input data for the five candidate local-ATAC-MVs. f An illustration of data trinarization and edit distance with multinomial distributions. The trinarization details can be found in Methods. g The edit distance variables quantify how many strains have 0, 1, or 2 absolute distances between \(\widetilde {\mathbf {Y_{g}}}\) and \(\widetilde {\mathbf {R_{g}}}\)/\(\widetilde {\mathbf {B_{g}}}\) and are modeled by multinomial distributions