Plasma Proteomic Profiles Predict Individual Future Osteoarthritis Risk

Area of Science:

• Biochemistry and Molecular Biology
• Genetics and Genomics
• Biomarker Discovery

Background:

• Osteoarthritis (OA) poses a significant socioeconomic burden, yet early diagnosis remains challenging due to a lack of sensitive biomarkers.
• Current diagnostic limitations hinder timely intervention and effective management of OA.

Purpose of the Study:

• To identify plasma proteins associated with the future risk of developing OA.
• To develop a predictive model for OA risk using proteomic and clinical data.

Main Methods:

• Large-scale proteomic analysis of 45,307 UK Biobank participants using Olink Explore Proximity Extension Assay (1,463 proteins).
• Construction of a predictive model with LightGBM machine learning and SHapley Additive exPlanations (SHAP) for variable importance.
• Integration of plasma protein data with clinical variables and electronic health records.

Main Results:

• Identified a panel of proteins significantly associated with incident OA risk.
• COL9A1 and CRTAC1 emerged as the strongest predictors (HR 1.54 and 1.65, respectively).
• Proteins COL9A1 and CRTAC1 levels deviated over a decade before OA onset, indicating early detection potential. Predictive model achieved AUCs of 0.72-0.82 for OA risk prediction at various time points and joint-specific predictions.

Conclusions:

• Plasma proteomics can predict future OA risk, enabling preemptive strategies.
• An integrated model of proteomic biomarkers and clinical data offers a promising tool for OA risk assessment.
• This approach may optimize OA management and enhance prevention efforts.