Simultaneous estimation of bi-directional causal effects and heritable confounding from GWAS summary statistics

Summary

This paper presents LHC-MR, a methodological advance for Mendelian Randomisation that addresses two critical limitations of existing approaches: under-exploitation of genome-wide markers and sensitivity to heritable confounding. By simultaneously estimating bi-directional causal effects and confounder contributions whilst accounting for sample overlap, the method revealed previously hidden causal mechanisms (such as disease diagnosis prompting lifestyle improvements) and identified protective effects (e.g. HDL cholesterol against systolic hypertension) obscured in standard analyses.

UK applicability

This is a statistical methodology paper of potential relevance to UK health research and epidemiology, particularly for institutions conducting genome-wide association studies or Mendelian Randomisation analyses. The improved causal inference framework could inform UK biobank research and evidence synthesis on diet-disease relationships, though direct agricultural or farming systems application is limited.

Key measures

Bi-directional causal effects, direct heritabilities, confounder effects, and performance metrics compared against existing Mendelian Randomisation methods across simulation settings and real trait associations

Outcomes reported

The study developed and validated the Latent Heritable Confounder Mendelian Randomisation (LHC-MR) method for estimating bi-directional causal effects between risk factors and complex human traits using genome-wide association study summary statistics. The method was applied to 13 complex traits to identify causal relationships and heritable confounders previously undetected by standard MR approaches.

Topic tags