Molecular Metabolism (Jul 2022)

Gene-based whole genome sequencing meta-analysis of 250 circulating proteins in three isolated European populations

  • Arthur Gilly,
  • Lucija Klaric,
  • Young-Chan Park,
  • Grace Png,
  • Andrei Barysenka,
  • Joseph A. Marsh,
  • Emmanouil Tsafantakis,
  • Maria Karaleftheri,
  • George Dedoussis,
  • James F. Wilson,
  • Eleftheria Zeggini

Journal volume & issue
Vol. 61
p. 101509

Abstract

Read online

Objective: Deep sequencing offers unparalleled access to rare variants in human populations. Understanding their role in disease is a priority, yet prohibitive sequencing costs mean that many cohorts lack the sample size to discover these effects on their own. Meta-analysis of individual variant scores allows the combination of rare variants across cohorts and study of their aggregated effect at the gene level, boosting discovery power. However, the methods involved have largely not been field-tested. In this study, we aim to perform the first meta-analysis of gene-based rare variant aggregation optimal tests, applied to the human cardiometabolic proteome. Methods: Here, we carry out this analysis across MANOLIS, Pomak and ORCADES, three isolated European cohorts with whole-genome sequencing (total N = 4,422). We examine the genetic architecture of 250 proteomic traits of cardiometabolic relevance. We use a containerised pipeline to harmonise variant lists across cohorts and define four sets of qualifying variants. For every gene, we interrogate protein-damaging variants, exonic variants, exonic and regulatory variants, and regulatory only variants, using the CADD and Eigen scores to weigh variants according to their predicted functional consequence. We perform single-cohort rare variant analysis and meta-analyse variant scores using the SMMAT package. Results: We describe 5 rare variant pQTLs (RV-pQTL) which pass our stringent significance threshold (7.45 × 10−11) and quality control procedure. These were split between four cis signals for MARCO, TEK, MMP2 and MPO, and one trans association for GDF2 in the SERPINA11 gene. We show that the cis-MPO association, which was not detectable using the single-point data alone, is driven by 5 missense and frameshift variants. These include rs140636390 and rs119468010, which are specific to MANOLIS and ORCADES, respectively. We show how this kind of signal could improve the predictive accuracy of genetic factors in common complex disease such as stroke and cardiovascular disease. Conclusions: Our proof-of-concept study demonstrates the power of gene-based meta-analyses for discovering disease-relevant associations complementing common-variant signals by incorporating population-specific rare variation.

Keywords