Multi-ancestry meta-analysis identifies 2 novel loci associated with ischemic stroke and reveals heterogeneity of effects between sexes and ancestries

Abstract

Cerebrovascular accident (stroke) is the second leading cause of death and disability worldwide. Stroke prevalence varies by sex and ancestry, which could be due to genetic heterogeneity between subgroups. We performed a genome-wide meta-analysis of 16 biobanks across multiple ancestries to study the genetic contributions underlying ischemic stroke (60,176 cases, 1,310,725 controls) as part of the Global Biobank Meta-analysis Initiative (GBMI). Two novel loci associated ischemic stroke with plausible candidate genes, FGF5 and CENPQ/MUT, were identified after replication in four additional datasets. One locus showed significant ancestry heterogeneity (PDE3A) and two loci showed significant sex-heterogeneity (SH3PXD2A and ALDH2). The ALDH2 locus had a male-specific association for stroke in GBMI (P-value males = 1.67e-24, P-value females = 0.126). To test whether we would see a difference in the predictive power of sex-specific polygenic risk scores (PRSs), we compared the C-indexes for sex-specific and sex-combined PRSs in HUNT dataset. A sex-combined PRS was more successful at predicting stroke cases than a sex-specific PRS, most likely due to more stable effect estimates from the sex-combined summary-statistics. These approaches can be applied to further unravel the genetic underpinnings of stroke and other complex diseases.

Meta-analysis fine-mapping is often miscalibrated at single-variant resolution

“…Motivated by successful validation of SLALOM performance, we investigated whether fine-mapping confidence and resolution were improved in the GBMI meta-analyses over individual biobanks. To this end, we used our fine-mapping results 16,17 of nine disease endpoints (asthma 64 , COPD 64 , gout, heart failure 73 , IPF 62 , primary open angle glaucoma 74 , thyroid cancer, stroke 75 , and venous thromboembolism 76 ) in BBJ 58 , FinnGen 20 , and UKBB 19 Europeans that also contributed to the GBMI meta-analyses for the same traits.…”

“…To directly compare with fine-mapping results from the GBMI meta-analyses, we used our fine-mapping results of nine disease endpoints (asthma 64 , COPD 64 , gout, heart failure 73 , IPF 62 , primary open angle glaucoma 74 , thyroid cancer, stroke 75 , and venous thromboembolism 76 ) in BBJ 58 , FinnGen 20 , and UKBB 19 Europeans that were also part of the GBMI meta-analyses for the same traits. For comparison, we computed the maximum PIP for each variant and the minimum size of 95% CS across BBJ, FinnGen, and UKBB.…”

Global Biobank Meta-analysis Initiative: Powering genetic discovery across human disease

“…Article ll OPEN ACCESS endpoints, [27][28][29]42,43 (2) to systematically characterize genomewide significant loci via fine mapping, 40 transcriptome-wide association, 35 protein QTL Mendelian randomization analysis, 36 and drug target prioritization, 44 and (3) to improve the disease risk prediction with PRSs based on the multi-biobank multiancestry meta-analysis results. 9 Together, the pilot work conducted in GBMI shows that biobanks can be meta-analyzed to provide reliable genetic discoveries despite the heterogeneous characteristics across biobanks in many aspects, such as locations, sample sizes, genotyping and phenotyping approaches, sample ancestries, and strategies to recruit participants, with standardized phenotype definitions and analysis pipelines.…”

Global biobank analyses provide lessons for developing polygenic risk scores across diverse cohorts

“…To boost statistical power, we can meta-analyze GBMI GWAS with other non-overlapping cohorts as shown in other GBMI working groups [33][34][35] . However, we should note that more heterogeneity might be introduced from different resources such as population structure and phenotype definitions, which we cannot control with summary statistics data and that could exacerbate the heterogeneous performance of PRS across target populations.…”