Multiple polygenic scores improve bone mineral density prediction in an independent sample of Caucasian women

Abstract

Purpose of the studyTo determine if multiple Genetic Risk Scores (GRSs) improve bone mineral density (BMD) prediction over single GRS in an independent sample of Caucasian women.Study designBased on summary statistics of four genome-wide association studies related to two osteoporosis-associated traits, namely BMD and heel quantitative ultrasound derived estimated BMD (eBMD), four GRSs were derived for 1205 individuals in the Genome-Wide Scan for Female Osteoporosis Gene Study. The effect of each GRS on BMD variation was assessed using multivariable linear regression, with conventional risk factors adjusted for. Next, the eBMD-related GRS that explained the most variance in BMD was selected to be entered into a multi-score model, along with the BMD-related GRS. Elastic net regularised regression was used to develop the multiscore model, which estimated the joint effect of two GRSs (GRS_BMD and GRS_eBMD) on BMD variation, after being adjusted for conventional risk factors.ResultsWith the same clinical risk factors having been adjusted for, the model that included GRS_BMD performed best by explaining 32.53% of the variance in BMD; the single-score model that included GRS_eBMD explained 34.03% of BMD variance. The model that includes both GRS_BMD and GRS_ eBMD, as well as the clinical risk factors, aggregately explained 35.05% in BMD variation. Compared with the single GRS models, the multiscore model explained significantly more variance in BMD.ConclusionsThe multipolygenic score model explained a considerable amount of BMD variation. Compared with single score models, multipolygenic score model provided significant improvement in explaining BMD variation.