KDRD: assessing discovery replicability based on summary association statistics from a single study

KDRD estimates expected replicability from summary association statistics in a single study, and in UK Biobank proteomics of 45 brain-related traits, 33 traits show at least moderate replicability.

• Why it matters: Replicability checks in large-scale association studies often need independent datasets or repeated splitting, which can be limited by data access, computation, and reduced sample sizes.
• What they did: KDRD models observed association statistics with Gaussian kernels plus a theoretical null to estimate replicability between independent studies, and it was validated in random subsampling simulations and applied to full UK Biobank data.
• The result: In matched-sample comparisons, proteomic associations had higher KDRD estimates than pQTL associations across all 45 traits, and KDRD sample-size calculations suggest about 200,000 participants for moderate and 500,000 for high replicability.

AI summary of the abstract · may contain errors · not medical advice · read the paper.