Multi-ancestry analysis of plasma protein levels influencing and responding to major depression liability

The development of novel-acting antidepressant medications with fewer side effects and sustained efficacy requires an in-depth understanding of the aetiology of major depressive disorder (MDD) across diverse populations. Here we used a Mendelian randomization (MR) framework to identify protein levels that influence MDD risk, and that respond to MDD liability in the general population. We use summary-level data from four major ancestral groups to evaluate the consistency of genetic associations and MR estimates across populations. We identified 17 proteins that are putatively causal for MDD, with evidence of differential effects across ancestries for five proteins, which we replicate in independent individual level data. We also identified widespread protein level changes in response to disease liability in the general population. We showed that such associations can appear ancestry-specific until differential power is accounted for, after which the vast majority of associations appear consistent across ancestral groups. The protein response to disease liability can be used to generate a proteomic risk score that is strongly predictive of prospective MDD incidence. Our results indicate that multi-ancestry Mendelian randomization improves power for ancestral groups with smaller sample sizes and will inform our understanding of disease aetiology if differential marginal effects across populations arising due to gene-environment interactions can be studied.