Predicting SARS-CoV-2-specific CD4+ and CD8+ T-cell responses elicited by inactivated vaccines in healthy adults using machine learning models

This is the authors' abstract. We don't add key points for this paper.

The ongoing evolution of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants highlights the importance of monitoring immune responses to guide vaccination strategies. Although neutralizing antibodies (NAbs) have garnered increasing attention, T-cells are crucial for conferring long-lasting immunity, especially their resilience against viral mutations. However, assessing T-cell responses clinically has been hindered by cost and complexity. In this study, we recruited a cohort of 134 healthy adults, who had been immunized with three doses of the SARS-CoV-2 inactivated vaccine. Cellular immunity elicited by a comprehensive array of overlapping peptides covering the entire sequence of the virus's structural proteins was assessed by intracellular cytokine staining (ICS). Additionally, a dataset including demographic information, routine blood indices, and immune cell indicators comprising 32 variables was collected. Multivariate analysis revealed age and days post-vaccination as key factors influencing the strength of the T-cell response. Importantly, random forest (RF) and classification and regression tree (CART) algorithms were employed, along with 8 easily accessible indicators to formulate predictive models for the SARS-CoV-2-specific CD4⁺ and CD8⁺ T-cell responses. Besides, these models demonstrated substantial accuracy (r > 0.9) in both the training and testing sets. Our findings offer an efficient and economical methodology for evaluating the T-cell reactions in healthy adults following inactivated SARS-CoV-2 vaccination, which is visualizable and easy to use, providing a novel strategy for assessing cellular immunity after vaccination.