Abstract
INTRODUCTION: Tumor neoantigens possess high specificity and immunogenicity, making them crucial targets for personalized cancer immunotherapies such as mRNA vaccines and T-cell therapies. However, experimental identification and evaluation of their immunogenicity are time-consuming, which limits the efficiency of vaccine development. METHODS: To address these challenges, we implemented two key strategies. First, we upgraded the TumorAgDB database by integrating publicly available neoantigen data from the past two years, resulting in TumorAgDB2.0. Second, we developed NeoTImmuML, a weighted ensemble machine learning model for predicting neoantigen immunogenicity. Using data from TumorAgDB2.0, we calculated the physicochemical properties of peptides and systematically evaluated eight machine learning algorithms via five-fold cross-validation. The top-performing algorithms - LightGBM, XGBoost, and Random Forest - were integrated into a weighted ensemble model. RESULTS: TumorAgDB2.0 (https://tumoragdb.com.cn) now contains 187,223 entries. Moreover, NeoTImmuML demonstrated strong generalization performance on both internal and external test datasets. SHAP feature importance analysis revealed that peptide hydrophilicity and length are key determinants of immunogenicity. DISCUSSION: TumorAgDB2.0 provides a comprehensive data resource for neoantigen research, while NeoTImmuML offers an efficient and interpretable tool for predicting neoantigen immunogenicity. Together, they offer valuable support for the design of personalized neoantigen vaccines and the development of cancer immunotherapy strategies.