A frequentist test of proportional colocalization after selecting relevant genetic variants

Abstract

Introduction: Colocalization analyses assess whether two traits share the same causal genetic variants in a single gene region. Bayesian enumeration colocalization tests are now routinely used in practice; for example, for genetic analyses in drug development pipelines. Frequentist proportional colocalization tests operate under markedly different assumptions, testing the null hypothesis that genetic associations with two traits are proportional, therefore can provide valuable complementary evidence in cases where enumeration colocalization results are inconclusive or sensitive to priors. Methods: We propose a novel conditional frequentist test of proportional colocalization, prop-coloc-cond, that accounts for uncertainty in both lead variant selection and a first-stage slope test, in order to recover accurate type I error control. The test can be implemented straightforwardly, requiring only summary data on genetic associations. Results: Simulation evidence demonstrates that prop-coloc-cond achieves competitive finite-sample type I error control relative to existing proportional colocalization tests. An empirical investigation of GLP1R gene expression illustrates specific scenarios in which proportional and enumeration colocalization approaches give concordant or discordant evidence, and how careful interrogation of discordant results can inform the most likely causal model. Conclusion: Since proportional and enumeration colocalization approaches differ in their underlying assumptions and maintained hypotheses, combining evidence from both approaches in the spirit of triangulation can yield more robust colocalization inferences than either method alone.