Fig. 3
From: Estimating colocalization probability from limited summary statistics
Scenarios that reduce POEMColoc performance. a POEMColoc-1 sensitivity to mismatched ancestry panels. Performance at identifying colocalization (\({H}_{4}\)) in datasets simulated under \({H}_{0}-{H}_{4}\). Data was simulated from European only ancestry individuals and the indicated population was used as a reference panel. b POEMColoc-2 sensitivity to mismatched ancestry panels. Performance at identifying colocalization (\({H}_{4}\)) in datasets simulated under \({H}_{0}-{H}_{4}\). c An example of a dataset in which there are multiple independent GWAS signals. Only the second peak is colocalized with the eQTL, so POEMColoc does not detect it. d Precision-recall curves for predicting colocalization using coloc (\(P({H}_{4})>0.9\)) in the UK Biobank where the POEMColoc top SNP was restricted to SNPs on different sized Illumina array genotyping subsets (1 M—OmniExpressExom8; 2.5 M—HumanOmni2.5Exome; 5 M—HumanOmni5Exome)