1036Independent evaluation of melanoma polygenic risk scores in UK and Australian prospective cohorts

Abstract Background To improve melanoma early detection, tools to predict personal risk based on genetic information (polygenic scores, PRS) have been developed, but require external validation. Methods We analysed invasive incidence in UK Biobank (UKB; n = 395,647; 1,651 cases) and the Melbourne Collaborative Cohort Study (MCCS, Australia; 4,765; 303 cases). Three PRS were evaluated: 68 variants (SNPs) at 54 loci from a 2020 meta-analysis (PRS68); 50 SNPs significant excluding UKB (PRS50); 45 21 known pre-2020 (PRS45). 10-year risks calculated population-level cancer registry data by age group sex, with without adjustment. Results All strongly associated incidence, including after adjustment for age, ethnicity, ease of tanning. Predicted absolute sex alone underestimated (ratio expected/observed cases E/O=0.65, 95% confidence interval 0.62-0.68) MCCS (E/O=0.65, 0.57-0.73). For UKB, this was reduced PRS-adjustment, e.g. PRS50-adjusted E/O=0.91 (0.87-0.95). Discriminative ability PRS68- higher than (deltaAUC 0.07-0.1, p < 0.0001), PRS45-adjusted 0.02-0.04, 0.001). Conclusions A derived larger, more diverse improves prediction compared an earlier PRS. Re-calibration may be necessary application specific populations. Key messages score can might help tailor prevention detection strategies different levels.