Here, we explored the feasibility to use a protein-based machine learning model for risk prediction of MetS in a prospective cohort.