VaxOptiML: Leveraging Machine Learning for Accurate Prediction of MHC-I & II Epitopes for Optimized Cancer Immunotherapy

Abstract

AbstractIn the realm of cancer immunotherapy, the ability to accurately predict epitopes is crucial for advancing vaccine development. Here, we introduce VaxOptiML (available athttps://vaxoptiml.streamlit.app/), an integrated pipeline designed to enhance epitope prediction and prioritization. Utilizing a curated dataset of experimentally validated epitopes and sophisticated machine learning techniques, VaxOptiML features three distinct models that predict epitopes from target sequences, pair them with personalized HLA types, and prioritize them based on immunogenicity scores. Our rigorous process of data cleaning, feature extraction, and model building has resulted in a tool that demonstrates exceptional accuracy, sensitivity, specificity, and F1-score, surpassing existing prediction methods. The robustness and efficacy of VaxOptiML are further illustrated through comprehensive visual representations, underscoring its potential to significantly expedite epitope discovery and vaccine design in cancer immunotherapy, Additionally, we have deployed the trained ML model using Streamlit for public usage, enhancing accessibility and usability for researchers and clinician.