Figure 3.

CAFEH GTEx and simulations

(A) A schematic representation of CAFEH. CAFEH can be viewed as a sparse regression with a shared set of causal variants across all studies. Entries of W are modeled with a spike and slab prior, so each study uses a subset of causal variants.

(B) Proportion of false discoveries (FDP) versus proportion of true positives (Power) across a range of colocalization thresholds.

(C) FDP versus Power across a range of thresholds of the posterior inclusion probability in competing fine-mapping methods.

(D) Enrichment of the top CAFEH variant of each gene in promoter (teal) and enhancer (red) elements in matched Roadmap cell-types relative to top eQTL variants.

(E) Redefining colocalization with allelic heterogeneity. Top: representation of colocalization in any or the top component. Stars represent causal variants in each study. Bottom: Proportion of tissue pairs colocalizing in any or top CAFEH components at a 0.5 threshold.