Prediction of melanoma incidence based on combination of genetic variants

Abstract

The occurrence of melanoma is a composite process that implicates the interaction of phenotypic, environmental, and genetic risk factors. We constructed genetic risk models, with the aim to assess their predictive performance on melanoma risk. Summary level data from the largest meta-analysis of genome-wide association studies for melanoma, up to date, were used for the construction of weighted genetic risk scores. We used six different p-value thresholds for genetic variants inclusion. We evaluated our genetic risk scores in 2,862 events of incident melanoma and 321,789 cancer-free controls from the UK Biobank, a prospective cohort study of 500,000 participants. Using AUCs, we compared the predictive ability of the different genetic risk scores. Genetic risk scores were strongly associated melanoma risk. Odds Ratios ranged from 1.478 to 1.528. The predictive ability of the genetic risk scores ranged from 0.6234 to 0.6328 showing a moderate performance. Our study suggests that when the p-value threshold for genetic variants inclusion become more tolerant, the prediction performance of the model improved. Validation of the results in larger populations, as well as Southern European populations is needed.