Extended Data Fig. 4: Performance of pgBoost trained on eQTL data from all tissues vs. whole blood.

Average enrichment across recall values of links predicted by pgBoost trained on fine-mapped eSNP-eGene pairs from all GTEx tissues vs. fine-mapped eSNP-eGene pairs from whole blood for (a) 4,434 fine-mapped eSNP-eGene pairs attaining maximum PIP > 0.5 across GTEx tissues, (b) 53,701 SNP-gene pairs attaining maximum ABC score > 0.2 across 344 biosamples, (c) 892 links validated by CRISPR, (d) 155 non-coding SNP-gene pairs derived from fine-mapped GWAS variants with a unique fine-mapped coding variant within a 2 Mb window, and (e) 712 fine-mapped eSNP-eGene pairs attaining PIP > 0.5 in GTEx whole blood, at various distance thresholds. The number of positive evaluation links at each distance threshold is specified in parentheses. Confidence intervals denote standard errors. Stars denote 2-sided bootstrap p-values for difference (*: p < 0.05, **: p < 0.01, ***: p < 0.001) of top method vs. each other method (Methods). For pgBoost_Blood-eQTL, we defined a training set of 1,518 eSNP-eGene pairs attaining PIP > 0.2 in whole blood as positives and 6,997 eSNP-eGene pairs including a gene in the positive set and attaining PIP < 0.01 in whole blood as negatives.