Proteomic Signatures for Risk Prediction of Atrial Fibrillation

Area of Science:

• Cardiovascular disease research
• Proteomics and biomarker discovery
• Predictive analytics in healthcare

Background:

• Proteomic signatures hold promise for improving disease prediction and enabling targeted interventions.
• The study investigates the utility of a protein risk score derived from large-scale proteomics data for predicting atrial fibrillation (AF).

Purpose of the Study:

• To determine if a protein risk score, developed from plasma proteomics data, can improve the prediction of incident atrial fibrillation (AF).

Main Methods:

• A protein risk score was developed using lasso-penalized Cox regression in a subset of 51,680 individuals from the UK Biobank Pharma Proteomics Project (UKB-PPP).
• The score was validated internally on a separate subset and externally in the Atherosclerosis Risk in Communities (ARIC) study.
• Performance was evaluated using C-index and risk reclassification metrics.

Main Results:

• The protein risk score, comprising 165 plasma proteins, significantly predicted incident AF in the UKB-PPP test set (HR 2.20 per 1-SD increase).
• The integrated model including the protein risk score demonstrated improved predictive accuracy (C-index 0.816) compared to models without it.
• The protein risk score led to a 5.4% improvement in risk reclassification at a 5-year risk threshold of 5% and showed consistent results in the ARIC study.

Conclusions:

• A protein risk score derived from a single plasma sample effectively improves the prediction of atrial fibrillation.
• These findings highlight the potential of proteomic signatures for AF screening and prevention, warranting further investigation.