Conditional-prevalence correction for liability-scale analyses of covariate-adjusted binary GWAS.

A Gaussian liability model yields a shared conversion factor for direct and collider-bias components when translating covariate-adjusted binary GWAS effects from log odds to liability scale.

• Why it matters: Covariate adjustment in heritable disease GWAS can induce collider bias, producing spurious genetic signals, and liability-scale interpretation needs a correct log-odds conversion.
• What they did: The work derives the liability-scale conversion for small genetic effects under a Gaussian liability model, characterizes its dependence on disease-risk variation across covariate values, and extends it to case-control sampling.
• The result: Simulations show the proposed conversion reduces effect-size inflation from scaling based only on disease prevalence, linking collider-bias corrections to direct liability-effect interpretation.

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