Plasma proteomics reveals clinical and mechanistic heterogeneity among individuals who develop coronary artery disease

The research included 42,803 participants from the UK Biobank, of whom 3,713 individuals developed coronary artery disease (CAD) within 10 years. Researchers identified a panel of 320 proteins from 2,923 baseline proteins that improved CAD prediction beyond clinical risk scores. Using reverse graph embedding, they reduced proteomic data to two dimensions and mapped each case onto a two-dimensional latent proteomic space. These dimensions were significantly associated with cardiometabolic and kidney-related clinical markers, with patterns replicated in the EPIC-Norfolk study. Analyses further linked these dimensions to 10-year incidence of various diseases including type 2 diabetes, obesity, and chronic kidney disease. Adding proteomic dimensions to clinical models improved prediction of 10-year chronic kidney disease incidence and other diseases. Pathway enrichment analyses revealed changes in extracellular matrix organization and immune programs among proteins contributing to the proteomic dimensions.