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

Abstract

SummaryWith the increasing availability of biobank-scale datasets that incorporate both genomic data and electronic health records, many associations between genetic variants and phenotypes of interest have been discovered. Polygenic risk scores (PRS), which are being widely explored in precision medicine, use the results of association studies to predict the genetic component of disease risk by accumulating risk alleles weighted by their effect sizes. However, limited studies have thoroughly investigated best practices for PRS in global populations across different diseases. In this study, we utilize data from the Global-Biobank Meta-analysis Initiative (GBMI), which consists of individuals from diverse ancestries and across continents, to explore methodological considerations and PRS prediction performance in 9 different biobanks for 14 disease endpoints. Specifically, we constructed PRS using heuristic (pruning and thresholding, P+T) and Bayesian (PRS-CS) methods. We found that the genetic architecture, such as SNP-based heritability and polygenicity, varied greatly among endpoints. For both PRS construction methods, using a European ancestry LD reference panel resulted in comparable or higher prediction accuracy compared to several other non-European based panels; this is largely attributable to European descent populations still comprising the majority of GBMI participants. PRS-CS overall outperformed the classic P+T method, especially for endpoints with higher SNP-based heritability. For example, substantial improvements are observed in East-Asian ancestry (EAS) using PRS-CS compared to P+T for heart failure (HF) and chronic obstructive pulmonary disease (COPD). Notably, prediction accuracy is heterogeneous across endpoints, biobanks, and ancestries, especially for asthma which has known variation in disease prevalence across global populations. Overall, we provide lessons for PRS construction, evaluation, and interpretation using the GBMI and highlight the importance of best practices for PRS in the biobank-scale genomics era.

Identifying and correcting for misspecifications in GWAS summary statistics and polygenic scores

“…We design GWAS simulations where variants have different sample sizes, which is often the case when meta-analyzing GWAS summary statistics from multiple cohorts with different sets of variants (Wang et al ., 2021). Using 40,000 variants from chromosome 22 (Methods), we simulate quantitative phenotypes with a heritability of 20% and 2000 causal variants.…”

“…Note that LD blocks are already widely used by several methods (such as lassosum and PRS-CS) because they allow for processing smaller matrices at once (Mak et al ., 2017; Ge et al ., 2019); here we have shown that these blocks are also useful to make methods more robust. PRS-CS is currently one of the most robust PGS methods; for example, it can use the (small) 1000 Genomes data as LD reference (Ge et al ., 2019), and can even use a European LD reference panel with multi-ancestry GWAS summary statistics (Wang et al ., 2021). We think this is made possible by the use of a (very) strong regularization in PRS-CS (, which would approximately correspond to using s = 0.5 in lassosum, δ = 1 in lassosum2, and 0.5 in LDpred2-auto).…”

“…European LD reference panel with multi-ancestry GWAS summary statistics (Wang et al, 2021). We think this is made possible by the use of a (very) strong regularization in PRS-CS (φ −1 ψ −1 j ≥ 1, which would approximately correspond to using s = 0.5 in lassosum, δ = 1 in lassosum2, and shrink_corr = 0.5 in LDpred2-auto).…”

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

“…The ischemic stroke PRS did not significantly predict stroke in the HUNT dataset when added on top of the age and sex information. Additionally, in an accompanying paper by Wang et al, the performance of the GBMI derived PRS was compared to one derived using the MegaStroke summary statistics (Wang et al, 2021). Wang et al concluded that the previous meta-analysis, with more cases underlying the summary statistics, performed better for African ancestry individuals whereas the GBMI derived PRS was slightly better for European individuals.…”

A practical guideline of genomics-driven drug discovery in the era of global biobank meta-analysis

“…Because of the diverse ancestry in the GWASs for meta-analysis, there was remarkable heterogeneity in the effective sample sizes across the genome-wide variants of the GBMI results of each phenotype. This heterogeneity affected the performance of downstream analyses, including those of polygenic risk score 65 . Therefore, we excluded variants with effective sample sizes <50% of the maximum effective sample size from the GWAS summary statistics of each phenotype.…”