UK Biobank release and systematic evaluation of optimised polygenic risk scores for 53 diseases and quantitative traits

Abstract

We present and assess the UK Biobank (UKB) Polygenic Risk Score (PRS) Release, a set of PRSs for 28 diseases and 25 quantitative traits being made available on the individuals in UKB. We also release a benchmarking software tool to enable like-for-like performance evaluation for different PRSs for the same disease or trait. Extensive benchmarking shows the PRSs in the UKB Release to outperform a broad set of 81 published PRSs. For many of the diseases and traits we also validate the PRS algorithms in other cohorts. The availability of PRSs for 53 traits on the same set of individuals also allows a systematic assessment of their properties, and the increased power of these PRSs increases the evidence for their potential clinical benefit.

GATTACA AS A LENS ON CONTEMPORARY GENETICS Marking 25 years into the film’s “not-too-distant” future

“…At the same time, generalizing genetic predictions of complex traits outside the settings in which they have been calibrated has proven difficult. For example, predictions from the massive height study mentioned above accounted for only 10-20% of the variance in height outside people of European ancestries (Yengo et al, 2022), a pattern that has been observed repeatedly in analogous contexts (Martin et al, 2017(Martin et al, , 2019Thompson et al, 2022). The causes of these problems with generalizability are not fully understood, but evidence points to contributions from worldwide differences in the frequencies and correlations of genetic variants (Wang et al, 2020) as well as genetic interactions (Mostafavi et al, 2020;Zhu, Ming, Cole, Kirkpatrick, & Harpak, 2022;Patel et al, 2022), in which the effect of a genetic variant on a trait depends on the environment (gene-by-environment interaction) or on genotypes at other positions in the genome (gene-by-gene interaction, or epistasis).…”

“…For the complex trait that has been most intensively studied, human height, a titanic meta-analysis of 5.4 million participants has led to a genetic predictor-i.e., a "polygenic score" (Torkamani, Wineinger, & Topol, 2018;Thompson et al, 2022)-that explains 40% of the variance in height in a sample of people of European ancestries (Yengo et al, 2022). This enormous effort achieved an accuracy approaching the "SNP heritability" (Yang, Zeng, Goddard, Wray, & Visscher, 2017;Hou et al, 2019), which is the portion of the heritability explained by common genetic variants.…”

Improving on polygenic scores across complex traits using select and shrink with summary statistics

“…As we have done here, comparing a particular PGS formation method to others across a variety of phenotypes and a variety of ancestries serves as a powerful benchmark of PGS model performance. Recently, a UK Biobank Polygenic Risk Score (PRS) method has been released as a resource of polygenic scores across many diseases and traits, with benchmarking of multiple PGS algorithms or published PGSs (Thompson et al, 2022) against this new method. Notably ovarian cancer was the only phenotype where the UKB generated PRS did not improve on the previously reported PRS, which we previously generated using the S4 method (Dareng et al 2022).…”

Contribution of genetics and lifestyle to the risk of major cardiovascular and thromboembolic complications following COVID-19

“…The development of the enhanced score used some UK Biobank participants, with the score then computed for those remaining to avoid the risk of overfitting. 40 In this study, we used the standard PRS for coronary artery disease (CAD), atrial fibrillation (AF), ischaemic stroke (ISS), and venous thromboembolic disease (VTE) in the primary analysis and the enhanced PRS in one of the sensitivity analyses. The continuous PRS used in this study was also categorized as high-risk (fifth quintile), intermediate-risk (2-4 quintiles), and low-risk (lowest quintile) to enable comparisons with previous studies.…”