Back

Multi-molecular scores map process-specific polygenic diabetes risk to atherosclerosis, cardiometabolic diseases, and vascular complications

Li, H.; Morze, J.; Adiels, M.; Kullberg, J.; Merino, J.; Sevilla-Gonzalez, M.; Bergström, G.; Orho-Melander, M.; Söderberg, S.; Östgren, C. J.; Lappalainen, T.; Fall, T.; Udler, M. S.; Häggström, I.; Wittenbecher, C.

2026-01-16 epidemiology
10.64898/2026.01.15.26344187 medRxiv
Show abstract

Type 2 diabetes (T2D) is etiologically heterogeneous, and process-specific polygenic scores (pPSs) only partly resolve this complexity. Here we integrate fourteen published T2D pPSs with plasma biochemistry, proteomics and NMR metabolomics in 29,425 SCAPIS participants to derive process-specific, polygenic-informed multi-molecular scores (pMMSs) that map genetic risk onto multi-omic molecular signatures. We then test the associations of these scores with subclinical coronary atherosclerosis, incident T2D, macrovascular disease, and microvascular complications in SCAPIS and 458,905 participants from UK Biobank. The novel pMMSs recapitulate the mechanistic interpretability of their underlying pPSs, yet show substantially stronger and more granular associations with molecular traits and disease risk. For example, per standard deviation, the Proinsulin pMMS indicates 4- to 5-fold increased risk of incident T2D and diabetic microvascular complications, whereas several lipid-related pMMSs highlight pleiotropic gene clusters with distinct roles in lipid metabolism and cardiometabolic disease pathogenesis. These process-specific molecular endophenotypes operationalize pleiotropy dissection for T2D risk and illustrate how polygenic risk propagates through molecular layers to shape complex cardiometabolic traits.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.