Polygenic scores via penalized regression on summary statistics

TLDR

Polygenic scores (PGS) quantify the genetic contribution to disease or phenotype, are used for risk stratification and as covariates, and recent interest focuses on methods that leverage published summary statistics, but the absence of linkage‑disequilibrium information in such statistics raises the question of how to incorporate external LD data. The authors propose lassosum, a penalized‑regression method that constructs PGS from summary statistics and a reference panel. They also introduce a general pseudovalidation approach for selecting the tuning parameter when validation phenotypes are unavailable. Simulations demonstrate that pseudovalidation yields prediction accuracy comparable to using a validation dataset, surpasses the conservative lowest‑parameter choice, and that lassosum outperforms clumping‑and‑thresholding in almost all scenarios while being faster and more accurate than LDpred.

Abstract

ABSTRACT Polygenic scores (PGS) summarize the genetic contribution of a person's genotype to a disease or phenotype. They can be used to group participants into different risk categories for diseases, and are also used as covariates in epidemiological analyses. A number of possible ways of calculating PGS have been proposed, and recently there is much interest in methods that incorporate information available in published summary statistics. As there is no inherent information on linkage disequilibrium (LD) in summary statistics, a pertinent question is how we can use LD information available elsewhere to supplement such analyses. To answer this question, we propose a method for constructing PGS using summary statistics and a reference panel in a penalized regression framework, which we call lassosum . We also propose a general method for choosing the value of the tuning parameter in the absence of validation data. In our simulations, we showed that pseudovalidation often resulted in prediction accuracy that is comparable to using a dataset with validation phenotype and was clearly superior to the conservative option of setting the tuning parameter of lassosum to its lowest value. We also showed that lassosum achieved better prediction accuracy than simple clumping and P ‐value thresholding in almost all scenarios. It was also substantially faster and more accurate than the recently proposed LDpred.