Performance of previously described polygenic risk scores for prostate cancer in a population with mixed ancestry.

Abstract

10578 Background: It is unclear whether polygenic risk scores (PRS) for prostate cancer (PrCa), developed using data from European populations, are applicable to individuals of mixed ancestry, such as the Brazilian population. We are conducting a case-control study to evaluate the performance of existing PrCa PRSs in Brazilian people. Methods: We prospectively included unselected PrCa patients (pt). The control group consisted of 347 elderly community men from Sao Paulo city (SABE cohort) included from 2000 to 2010, with whole genome sequencing (WGS 30x depth) data previously reported in the ABRAOM study. Men who developed cancer in the follow-up were excluded. The mean age at admission to the study was 72 years. DNA was extracted from blood samples for performing WGS at 15x depth (PrCa pt), using Illumina technology. VCF quality filtering and pre-processing were performed according to best practices of GATK, BCFtools and Plink. Ancestry composition was calculated using ancestry-informative markers derived from global populations. AUCs were calculated using the results from the total PRS value obtained with PGSCalc for cases and controls, without filters. We calculated PRS scores for 11 different PRSs for PrCa, described in 7 studies, derived from people with different ancestries and available in The Polygenic Score (PGS) Catalog or the literature (BARCODE1). Results: A total of 588 PrCa patients were included. The median age at diagnosis was 64.6 years. Most patients were diagnosed with clinical stages II (32.8%) and III (43.1%), and ISUP grades 2 (40.2%) and 3 (22.9%). Twenty PrCa pt, who were carriers of pathogenic or likely pathogenic germline variants on high and moderate penetrance prostate cancer predisposition genes, were excluded. The mean ancestry genetic composition for PrCa pt was 61.4% European and 24.6% African. We calculated the AUC for the Brazilian cohort (BZL) for 11 PRSs available in PGS Catalog. 1) PGS000030 - Schumacher et al. 2018; 147 SNPs - AUC BZL: 0.666; 2) PGS000751 - Du et al. 2019; 178 SNPs - AUC BZL: 0.694; 3) PGS000662 - Conti et al. 2021, 269 SNPs - AUC BZL: 0.705; 4) PGS002796, PGS002797, PGS002798, PGS002799 - Shi Z, et al. 2022; 232, 67, 128, 138 SNPs - AUC BZL, respectively: 0.684; 0.663; 0.670; 0.627; 5) PGS002240, PGS002241 - Mars et al. 2022; 1,092,093 (AUC BZL: -) and 6,497,734 SNP (AUC BZL: -) PRS for risk prediction of common diseases. 6) PGS003460 - Chen F et al. 2023; 278 SNPs - AUC BZL: 0.702; 7) Barcode1 - 130 SNPs - AUC BZL:0.674. Conclusions: We used WGS 15x to improve SNP capture associated with population genetic differences. PRS 269 (Conti et al., 2021) and PRS 278 (Chen F et al. 2023) yielded the best results, suggesting that a multiancestry-derived PRS may be applicable to the admixed Brazilian population. We continue including participants to confirm these results.