Extended Data Fig. 2: Evaluation of GeneBayes, related to Figs. 2 and 3.

a) Comparison of GWAS and LoF burden test associations for MCH. We took the minimum GWAS p-value within an LD block, and the minimum LoF burden test p-value for any gene that overlaps the LD block. Dotted lines indicate p = 5 × 10⁻⁸ for GWAS and p = 5 × 10⁻⁴, which corresponds to an FDR of 0.1 for the LoF burden test. Each dot corresponds to the LD block. Numbers of blocks in each quadrant are depicted on the top right corner. P-value is from a two-sided Fisher’s exact test. b) Correlation of burden test γ with All of Us. Plots are for the top 200 genes ranked by absolute values of raw γ (left) or GeneBayes posterior (right). c) Correlation of burden test γwith All of Us with different prior information. The result is for MCH. We ranked the genes based on absolute burden test effect size in UKB, either with or without applying GeneBayes with various patterns of prior information. d) Enrichment of GO and MsigDb hallmark pathways to top hits for MCH. The enrichment of the top 200 genes from the LoF burden test and GWAS is compared. For GWAS, the closest genes to the lead hits were ordered by p-values. For the LoF burden test, whether or not GeneBayes was applied, genes were ordered by absolute effect sizes. The GeneBayes posterior from various patterns of priors is also compared. e) Enrichment of top 200 genes from GWAS or LoF burden test with or without applying GeneBayes to representative pathways. The result is for MCH. f) Regulator-burden correlation for MCH is compared with their γ for MCH. Same comparison with Fig. 3c, but this time using γ before applying GeneBayes. Dotted lines indicate the same threshold with Fig. 3c. g) Correlation significance of HBA1 regulatory effects with gene effects across a variety of traits. Same comparison with Extended Data Fig. 3b, but this time using γ before applying GeneBayes. Dotted line indicates the same threshold with Extended Data Fig. 3b.