Can AI Design Cancer Vaccines? Evaluating Neural Networks for Epitope Prediction

Abstract

Immunotherapy as a form of cancer treatment can be effective, but often causes the immune system to attack healthy tissues, leading to significant side effects. Therapeutic cancer vaccines offer a safer, tumour-specific alternative, but their efficiency relies on accurate epitope prediction, which is used to identify regions of a protein that can trigger an immune response in a patient. This study evaluates MHCflurry on a clinically relevant melanoma-associated antigen to assess the real-world applications of computational epitope prediction to therapeutic melanoma vaccines. We assessed predicted epitopes based on binding affinity, presentation, and processing scores, identifying the peptide sequence AQAPATEEQEA as the strongest candidate. We visualized results and key findings for a quantitative analysis of the peptide sequences. Our findings suggest that while computational tools like MHCflurry show promise in the design of cancer vaccines, they require experimental validation before implementation or clinical application.