Human tumors frequently harbor multiple somatic mutations, generating neoantigens ideal for T-cell-based cancer immunotherapy. To identify these neoantigens for personalized vaccines, whole-exome sequencing (WES) of matched tumor and normal-cell DNA is performed to detect somatic mutations. These mutations are prioritized using RNA-Seq profiling, in silico prediction of their binding affinity to the major histocompatibility complex (MHC), and proteomic validation. However, existing analysis packages are fragmented, computationally intensive, and need more effective integration. Additionally, current in silico predictions of mutant peptide-MHC binding affinity lack accurate 3-dimensional modeling confirmation [1], which motivates us to develop an AI-enhanced workflow for Cancer Neoantigen Prioritization. This proposal will leverage NVIDIA Parabricks to accelerate somatic mutation identification, including genome alignment and variant calling. We will employ AlphaMissense [2], an AlphaFold adaptation refined with human and primate variant databases, to predict the pathogenicity of missense variants. Unlike NetMHCpan, we will incorporate TFold, an AlphaFold-based pipeline for peptide-MHC structure modeling [3], to provide mechanistic explanations for binding predictions. By leveraging multi-omics data from the same patient in the cancer research cohort collected by CGU and NHRI, we can prioritize high-confidence cancer neoantigens through proteomic validation. We will provide the FASTA sequences of candidate neoepitopes to our partners at NHRI for vector design and downstream immunotherapy research. Suppose the proposed workflow accurately predicts tumor-specific neoantigens and elicits anti-tumor T cell immunity. In that case, it will identify crucial targets for personalized cancer vaccines and adoptive T-cell therapies, significantly advancing personalized medicine.