An interpretable machine learning model integrating mucosal and cellular immunity to predict SARS-CoV-2 reinfection: a multicenter cohort study.

SARS-CoV-2 reinfection remains a significant public health challenge that requires accurate risk assessment across populations. Using prospectively collected data from two independent cohorts, we developed and externally validated an interpretable machine learning model for predicting symptomatic SARS-CoV-2 reinfection. The cohort comprised 1,156 adults from Wuxi and Nantong, China, with a median follow-up duration of 10 months. Reinfection occurred in 128 participants (11.1%). The model integrated eighteen clinical and immune features, and ultimately identified spike-specific nasal secretory immunoglobulin A (sIgA) and spike-specific CD4+ central memory T cells (TCM) as key predictors. Developed using the extreme gradient boosting (XGB) algorithm, the prediction model achieved an AUC of 0.835 in the internal validation cohort and demonstrated strong discrimination in the external validation cohort. SHAP analysis revealed that higher nasal spike-specific sIgA was associated with lower risk, while spike-specific CD4+ TCM cells exhibited a non-linear association with reinfection. This interpretable immune-based model provides a framework for immune-informed risk stratification, enabling better prediction and management of SARS-CoV-2 reinfection risk in settings with heterogeneous population immunity.