Figure 1.

Overview of development and validation of population-specific and multi-ancestry PRS for venous thromboembolism.

PRS development consisted of two steps: training ancestry-specific PRS and tuning multi-ancestry PRS. We trained ancestry-specific PRSs using European- and African ancestry GWAS summary statistics from the INVENT consortium and two Bayesian methods (LDPRED2 and PRSCSx). We then tuned the constructed multi-ancestry PRSs by regressing VTE case-control status on a linear combination of the two ancestry-specific PRSs in two separate tuning samples: one European-ancestry tuning sample and one African-ancestry tuning sample. NHS, Nurses’ Health Study; HPFS, Health Professional Follow-up Study; MGI, Michigan Genomics Initiative; UCLA, UCLA Precision Health Biobank; PMBB, Penn Medicine Biobank.