Assessing Type 2 Diabetes and GLP-1 agonist response trajectories with a proteogenomic atlas of disease progression

Type 2 diabetes (T2D) progresses through heterogeneous pathways that glycemic staging alone does not resolve. We constructed the Metabolic Atlas of the Proteome in Diabetes (MAP-D), leveraging Olink measurements of 2,923 circulating proteins in ~42,000 UK Biobank participants to map associations with three cardiometabolic hallmarks - adiposity (BMI), a proxy for insulin resistance (triglyceride-to-HDL cholesterol ratio), and glycemia (HbA1c) - across normoglycemia, prediabetes, and incident T2D. We triangulated cross-sectional associations with bidirectional Mendelian randomization and semaglutide trial proteomics to infer causal directionality. This revealed three distinct causal architectures: adiposity predominantly reshapes the proteome, glycemia is driven by upstream proteins, and insulin resistance shows bidirectional feedback. Integration with trial data identified proteins reversed by therapy and a subset of persistent proteins that remain dysregulated despite GLP-1RA treatment. These persistent proteins are associated with incident coronary artery disease and overlap with targets of approved therapies, nominating candidates for combination strategies beyond GLP-1RA monotherapy.