A prospective cohort-based artificial intelligence evaluation system for the protective efficacy and immune response of SARS-CoV-2 inactivated vaccines

• Published in: International immunopharmacology 2024-06, Vol.134, p.112141, Article 112141
• Main Authors: Zhang, Jin, Meng, Yuan, Yang, Mei, Hao, Wudi, Liu, Jianhua, Wu, Lina, Yu, Xiaojun, Zhang, Yue, Lin, Baoxu, Xie, Chonghong, Ge, Lili, Zhijie Zhang, Tong, Weiwei, Chang, Qing, Liu, Yong, Zhang, Yixiao, Qin, Xiaosong
• Format: Article
• Language: English
• Subjects: Artificial intelligence evaluation system; Cellular immunity; Humoral immunity; Prospective cohort
• ISSN: 1567-5769; 1878-1705; 1878-1705
• DOI: 10.1016/j.intimp.2024.112141

•The prospective cohort population in this study included 2,863 volunteers who underwent the COVID-19 inactivated vaccine booster strategy and 1,807 identified cases within the SARS-CoV-2 breakthrough cohort, and the population was monitored at different time points before and after infection breakthrough.•Booster shots boosted neutralising antibodies to 298.02 AU/mL within 2 weeks, and sex, age, disease history, and smoking status impacted post-booster antibody levels.•This study demonstrates the critical role of the synergistic effect of humoral immunity, cellular immunity, and epidemiological factors in determining the protective efficacy of COVID-19 vaccines post-booster administration. Background: Novel coronaviruses constitute a significant health threat, prompting the adoption of vaccination as the primary preventive measure. However, current evaluations of immune response and vaccine efficacy are deemed inadequate. Objectives: The study sought to explore the evolving dynamics of immune response at various vaccination time points and during breakthrough infections. It aimed to elucidate the synergistic effects of epidemiological factors, humoral immunity, and cellular immunity. Additionally, regression curves were used to determine the correlation between the protective efficacy of the vaccine and the stimulated immune response. Methods: Employing LASSO for high-dimensional data analysis, the study utilised four machine learning algorithms—logistical regression, random forest, LGBM classifier, and AdaBoost classifier—to comprehensively assess the immune response following booster vaccination. Results: Neutralising antibody levels exhibited a rapid surge post-booster, escalating to 102.38 AU/mL at one week and peaking at 298.02 AU/mL at two weeks. Influential factors such as sex, age, disease history, and smoking status significantly impacted post-booster antibody levels. The study further constructed regression curves for neutralising antibodies, non-switched memory B cells, CD4+T cells, and CD8+T cells using LASSO combined with the random forest algorithm. Conclusion: The establishment of an artificial intelligence evaluation system emerges as pivotal for predicting breakthrough infection prognosis after the COVID-19 booster vaccination. This research underscores the intricate interplay between various components of immunity and external factors, elucidating key insights to enhance vaccine effectiveness. 3D modelling discerned distinctive interacti