Large-scale multi-omics enhance risk prediction for type 2 diabetes

Area of Science:

• Genomics and Precision Medicine
• Biomarker Discovery
• Metabolomics and Proteomics

Background:

• Established clinical models for type 2 diabetes risk prediction have limitations.
• Polygenic risk scores (PRS), metabolomics, and proteomics show individual promise for improving prediction.
• The combined utility of these multi-omics biomarkers with clinical data is not well understood.

Purpose of the Study:

• To evaluate if integrating multi-omics biomarkers enhances 10-year type 2 diabetes risk prediction.
• To compare the predictive performance of multi-omics integration against single-omics extensions and the Cambridge Diabetes Risk Score (CDRS).

Main Methods:

• Analysis of 42,840 UK Biobank participants without diagnosed diabetes.
• Development of multi-omics prediction models by adding PRS, 11 metabolites, and 15 proteins to the CDRS.
• Evaluation of model performance using Harrell's C-index and net reclassification index (NRI) in derivation and validation sets.

Main Results:

• Proteomics alone significantly improved the C-index from 0.862 (CDRS) to 0.884.
• The full multi-omics model further increased the C-index to 0.891, outperforming the proteomics-only model.
• Significant improvements in predictive performance were observed with multi-omics integration.

Conclusions:

• Integrating proteomics, metabolomics, and diabetes PRS substantially improves type 2 diabetes risk prediction beyond single-omics.
• Selected proteins and metabolites link diabetes to cardiovascular disease pathways.
• While multi-omics offers significant gains, proteomics-only extension presents a simpler clinical translation pathway pending further validation.