A computational framework for deep learning-based epitope prediction by using structure and sequence information

摘要: Precise prediction of the antigens that bind to major histocompatibility complex (MHC) molecules is one most important steps for many immune response-related studies including identification neoantigen capable recognizing cancer cells implementation personalized immunotherapy as well vaccine/therapeutic protein development. To overcome limitations limited performance existing models, we developed a new convolutional neural network (CNN)-based epitope model by incorporating features sequence vectors (Prot2Vec), chemical and structural characteristics an MHC molecule. Our optimized achieve better than conventional methods when validated with benchmark dataset from Immune Epitope Database (IEDB). Along model, also web-based framework, called Kepre, provide interfaces our novel CNN-based functionalities. Kepre provides user-friendly automated interface prediction. Upon researcher's submission interest, predicts binding probability between each type molecules. runs analysis on high-throughput computing infrastructure results in user-interactable GUI pdf formatted report file.