TEIP: A Compact, Open-Source Framework for Predicting Tumor Epitope Immunogenicity in Glioblastoma Cancer Using Deep Learning and Multi-Modal Biological Features

Identifying immunogenic tumor epitopes (neoantigens) is critical for cancer vaccine and T-cell therapy design. We propose **TEIP** (Tumor Epitope Immunogenicity Predictor), an integrative computational pipeline that combines sequence-based deep learning, HLA-binding affinity prediction, and immunogenomic filters to predict and prioritize T-cell epitopes. TEIP uses a multi-branch neural network to model peptide HLA presentation and T-cell recognition, augmented with features like predicted binding affinity and gene expression. On benchmark datasets, TEIP achieves significantly improved prediction performance (e.g., area-under-ROC > 0.89) com- pared to baseline tools. We demonstrate that TEIPs prioritized neoantigens show broad patient coverage, high tumor-specific expression, and include challenging targets (e.g., cancer-testis and viral antigens) often missed by prior methods. These results highlight TEIPs potential to streamline neoantigen discovery for personalized immunotherapy.