Preprint

Genome-wide LD scores enable robust heritability and genetic correlation estimation from summary statistics across genetic ancestries

PreprintCode

Presented as a platform (oral) talk at the ASHG 2026 Annual Meeting, Montréal

Two genomes drawn as lines with arcs between correlated variants: a single-ancestry genome has only short arcs, an admixed genome also has long arcs that fall outside the window used by standard LD scores.
Standard LD scores sum correlations inside a window. Admixed genomes carry correlation between distant variants, which genome-wide LD scores include.

SUMMIT combines genome-wide LD scores, estimated efficiently with random projections, with GWAS summary statistics to estimate genome-wide and annotation-specific heritability and genetic correlation. In simulations across single-ancestry and admixed cohorts it was more accurate and less biased than prior methods, with the largest gains in admixed participants; across 40 traits it recovered 26 genetic correlations at FDR < 0.05 in the two smaller cohorts, where cov-LDSC found two. The method now runs on the All of Us Research Program, an admixed cohort drawn from across the United States.

All publications