Fig. 7
From: INFIMA leverages multi-omics model organism data to identify effector genes of human GWAS variants
INFIMA outperforms alternatives for fine-mapping DO mouse eQTLs. a Histogram of numbers of local-ATAC-MVs around DO-eQTL markers with window size of W=1 Mb. b Boxplot of INFIMA posterior probabilities of association, \(\widehat {V}_{g}\), across all genes. Red dashed line depicts the posterior probability cutoff for FDR of 0.05. c Evaluation of fine-mapping strategies with empirical cumulative distribution of normalized easy Hi-C scores. “Most Likely" and “Least Likely" refer to most and least likely predictions from INFIMA, respectively. d Boxplots depict the proportion of the candidate causal local-ATAC-MVs that are included in the credible set by INFIMA stratified by the size of candidate sets. The intervals on the x-axis are from quantile bins (20%, 40%, 60%, 80%, 100% percentiles) of number of local-ATAC-MVs around the eQTL marker, pg. Median values are displayed on each boxplot. e Proportion of times each of the individual multi-omic components are the leading contributors to the INFIMA prior probability of causality: correlation between ATAC-seq signal and founder eQTL effect sizes |cor(A,E)|, 0.207; correlation between ATAC-seq signal and founder gene expression |cor(A,B)|, 0.330; footprint, 0.277; and distance, 0.186. f The rank scores of the inferred causal local-ATAC-MVs when individual components are the top ranking contributors. The higher the rank scores are, the more INFIMA weights in the component when inferring causal local-ATAC-MVs.