Abstract
Pleural mesothelioma (PM) is a lethal cancer primarily caused by asbestos exposure. Not all exposed individuals develop PM, suggesting the involvement of additional factors. This underscores the need for robust predictive models integrating biomarkers from multi-omic domains to improve risk stratification and early detection. We developed and evaluated polygenic risk scores (PRS) and methylation risk scores (MRS) using a retrospective case-control study (749 participants: 387 PM cases, 362 controls) and a nested case-control European Prospective Investigation into Cancer and Nutrition (EPIC)-Meso study (268 participants: 134 preclinical PM cases, 134 matched controls) within the EPIC cohort. Genome-wide association analyses in the retrospective case-control study identified PM-associated variants. The PRS (1123 SNPs with p < 0.001) in the retrospective training subset stratified disease risk in the test set (ORs 3.46-9.54 across top percentiles) and improved model discrimination (AUC = 0.75 vs. 0.71 in baseline model, p = 0.04). In EPIC-Meso, PRS performance was limited (AUC = 0.52). External validation in the UK-Biobank (UKBB) confirmed a modest but consistent association with PM-risk. A Meta-PRS derived from the UKBB-FinnGen meta-analysis replicated this trend in the full retrospective dataset, showing higher OR across top percentiles (2.5-12.3) and improved discrimination (AUC 0.74 vs. 0.72, p = 0.016). MRS, with 68 differentially methylated CpGs (effect-size >|0.10|, FDR p < 0.05) in the retrospective training set, increased the AUC from 0.66 to 0.85 (p < 0.001) in the test set and from 0.51 to 0.62 in EPIC-Meso. PRS was most predictive in low-exposure groups, while MRS remained robust across exposure levels. Combined PRS-MRS models improved discrimination. Integrating multi-omic biomarkers can enhance PM-risk stratification and support earlier, targeted interventions in high-risk asbestos-exposed groups.