Profiling the Onco-metabolic Nexus and Improving Cancer Risk Prediction Performance: A Large-scale Cohort and Genome-Wide Pleiotropic Analysis

Early screening and targeted intervention can effectively reduce cancer burden. However, most studies have proposed polygenic risk score (PRS) to perform risk prediction and population risk stratification. In this study, we investigate the causal links and shared genetics between cancer and metabolic traits to map the onco-metabolic nexus. Using multivariable Cox models, we assessed 240 trait-cancer associations. Genomic analyses included genome-wide and local genetic correlations and genomic structural equation modeling (gSEM) to identify pathways linking metabolic traits to cancer. We then used Deterministic Bayesian Sparse Linear Mixed Model (DBSLMM) to build both single and integrative PRS models based on gSEM and compared their predictive performance. Most of the metabolic traits are risk factors to cancer, such as waist-hip ratio-colorectal cancer [hazard ratio (HR) = 1.34; 95% confidence interval (CI) = 1.23-1.46; P = 1.59 × 10-11]. In the genetic correlation analysis, we identified 41 significant pairs in the onco-metabolic nexus and 405 significant genomic regions. Mendelian randomization analysis revealed 17 significant causal pairs. The integrative PRS model combining gSEM for metabolic traits improved prediction, with 13.95% variance explained in kidney cancer. The study highlights the intertwined genetic and clinical relationships between cancers and metabolic traits, improving cancer screening and intervention. SIGNIFICANCE: By uncovering the shared genetic basis of cancer and metabolic traits, our work enables a novel integrative risk prediction model, which promises to enhance precision screening and targeted prevention strategies for high-risk populations.