Integrative Genomic and Proteomic Approaches to Thrombotic Events: Identifying Shared and Distinct Mechanisms and Enhancing Risk Assessment

Abstract

Leveraging data from 37,564 UK Biobank Pharma Proteomics Project participants, we investigated mechanisms and developed risk prediction models for thrombotic events, including ischemic stroke (IS), myocardial infarction (MI), and venous thromboembolism (VTE), within a unified multi-phenotype framework. We conducted proteome-wide association studies for each thrombotic event and a composite endpoint (first occurrence of IS, MI, or VTE), identifying 788 proteins associated with the composite and 325, 419, and 303 proteins associated with IS, MI, and VTE, respectively (FDR-adjusted P < 0.05). Mendelian randomization analyses supported putative causal relationships for 22 proteins, of which 12 were further corroborated by colocalization. Analysis of protein-protein interactions among thrombotic event-associated proteins highlighted core mediators, including inflammatory and leukocyte-trafficking factors (e.g., IL6, TNF, CXCL10, ICAM1, ITGAM) and vascular remodeling or angiogenic factors (e.g., VEGFA). Protein co-expression networks constructed from 2,919 proteins revealed 6 modules significantly associated with thrombotic events, and pathway enrichment analyses indicated both shared and phenotype-specific biology. For prediction, we integrated proteomic and conventional predictors in survival models, using LASSO for embedded selection of proteins. Across outcomes, combined (conventional + proteomics) models outperformed conventional predictors models (C-index improvement ΔC = 0.022 for IS, 0.020 for MI, 0.014 for VTE, and 0.027 for the composite; all P < 0.01). These findings advance mechanistic understanding, nominate druggable targets, and demonstrate improved risk prediction for thrombotic events.