MHCSeqNet: A deep neural network model for universal MHC binding prediction

Abstract

Immunotherapy is an emerging approach in cancer treatment that activates the host immune system to destroy cancer cells that express unique peptide signatures (neoepitopes). Administrations of cancer-specific neoepitopes in the form of synthetic peptide vaccine have been proven effective in both mouse models and human patients. Because only a tiny fraction of cancer-specific neoepitopes actually elicit immune response, selection of potent, immunogenic neoepitopes remains a challenging step in cancer vaccine development. A basic approach for immunogenicity prediction is based on the premise that effective neoepitope should bind with the Major Histocompatibility Complex (MHC) with high affinity. In this study, we developed MHCSeqNet, an open-source deep learning model, which not only competes favorably with state-of-the-art predictors but also exhibits promising generalization to new MHC class I and class II alleles. MHCSeqNet employed neural network architectures developed for natural language processing to model amino acid sequence representations of MHC allele and epitope peptide as sentences with amino acids as individual words. This consideration allows MHCSeqNet to accept new MHC alleles as well as peptides of any length. The flexibility offered by MHCSeqNet should make it a valuable tool for screening effective neoepitopes in cancer vaccine development.Availability and implementation https://github.com/cmbcu/MHCSeqNet

Current challenges for epitope-agnostic TCR interaction prediction and a new perspective derived from image classification

“…Using this type of interaction map as the model input differs from the prevailing approach seen in sequence-based prediction models, not only in the field of TCR-epitope prediction, but also for the modeling of other molecular interactions. The interacting molecules are usually supplied to a model as separate or concatenated inputs (1, 3,5,6,15,18,20,25,26). Those types of models need to learn an internal representation for each molecule separately, before being combined again in deeper layers.…”

MHCAttnNet: predicting MHC-peptide bindings for MHC alleles classes I and II using an attention-based deep neural model

“…We use the IEDB ( Sahin et al , 2017 ) (as of 2019) to prepare our training and testing data as done in previous works ( Hu et al , 2018a ; Jurtz et al , 2017 ; Nielsen and Lund, 2009 ; Phloyphisut et al , 2019 ; Zeng and Gifford, 2019b ). We filter the data-points that correspond to ‘human’ or ‘homo-sapiens’ and the data-points that have MHC alleles belonging to classes I or II.…”

“…The PUFFIN model predicts the expected affinity of an MHC-peptide binding, for both classes I and II alleles, as well as the uncertainty of its prediction. The MHCSeqNet ( Phloyphisut et al , 2019 ) model considers the input as an equally-weighted linear amino acid chain and can handle 92 alleles of class I. It looks at all amino acid subsequences equally which is not necessary, as only certain subsequences contribute to determining the binding affinity.…”

A Pan-Specific GRU-Based Recurrent Neural Network for Predicting HLA-I-Binding Peptides

“…Phloyphisut et al 17 also developed a pan-specific MHCSeqNet model for predicting MHC-binding peptides. In this paper, the peptide sequences were embedded by a 1-gram model and then processed by a BiGRU layer.…”