Integrative proteogenomic and observational analysis identifies potential biomarkers for latent autoimmune diabetes in adults.

Summary

Proteome-wide MR and colocalization linked higher genetically predicted plasma CXCL10, SAA1, and SAA2 to increased LADA risk. CXCL10 showed the most robust evidence and demonstrated strong discrimination of LADA from T2D and healthy controls in a matched clinical cohort (ROC-AUC 0.838–0.889) with good calibration. Phenome-wide MR suggested no major safety concerns regarding CXCL10 as a biomarker/target.

Key Findings

• Proteome-wide MR identified CXCL10, SAA1, and SAA2 as proteins associated with increased LADA risk; CXCL10 had the strongest support.
• Colocalization, multi-tissue eQTL, and multivariable MR analyses reinforced the CXCL10–LADA association.
• In a matched clinical cohort (n=241), CXCL10 discriminated LADA from T2D and controls with ROC-AUC 0.838–0.889 and good calibration.
• Phenome-wide MR across 1,006 traits revealed no major safety concerns related to CXCL10 as a biomarker/target; druggability assessment supports ongoing target exploration.

Clinical Implications

CXCL10 testing could improve early LADA recognition and guide appropriate therapy (e.g., earlier insulin, autoimmunity-focused management), but requires multi-ancestry, prospective validation before routine adoption.

Limitations

• MR discovery largely in European ancestry; observational validation in a single-center Chinese cohort
• Diagnostic cut-offs and clinical utility require prospective, multi-ethnic validation and head-to-head comparison with existing algorithms

Future Directions

Prospective, multi-ancestry validation of CXCL10-based diagnostic algorithms for LADA; evaluation of longitudinal dynamics and integration with islet autoantibodies and genetic risk scores.