NetMHCpan-3.0; improved prediction of binding to MHC class I molecules integrating information from multiple receptor and peptide length datasets

Abstract

BackgroundBinding of peptides to MHC class I molecules (MHC-I) is essential for antigen presentation to cytotoxic T-cells.ResultsHere, we demonstrate how a simple alignment step allowing insertions and deletions in a pan-specific MHC-I binding machine-learning model enables combining information across both multiple MHC molecules and peptide lengths. This pan-allele/pan-length algorithm significantly outperforms state-of-the-art methods, and captures differences in the length profile of binders to different MHC molecules leading to increased accuracy for ligand identification. Using this model, we demonstrate that percentile ranks in contrast to affinity-based thresholds are optimal for ligand identification due to uniform sampling of the MHC space.ConclusionsWe have developed a neural network-based machine-learning algorithm leveraging information across multiple receptor specificities and ligand length scales, and demonstrated how this approach significantly improves the accuracy for prediction of peptide binding and identification of MHC ligands. The method is available at www.cbs.dtu.dk/services/NetMHCpan-3.0.Electronic supplementary materialThe online version of this article (doi:10.1186/s13073-016-0288-x) contains supplementary material, which is available to authorized users.

Identification of immediate early gene products of bovine herpes virus 1 (BHV-1) as dominant antigens recognized by CD8 T cells in immune cattle

“…This method is trained on an extensive set of close to 140 000 MHC-peptide binding values covering 158 MHC molecules, including 13 bovine MHC class I (BoLA-I) alleles, and allows for prediction of peptides capable of binding to any MHC class I molecule of known protein sequence. NetMHCpan version 2.9 [30] was used rather than the recently updated version 3.0 [54], as the former was trained on an extended set of BoLA-I binding data that are not available in the public-domain data used to train version 3.0.…”

The Prediction and Function of Neoantigen in Oncobiology

“…So, great efforts have been paid to develop the algorithms to accurately predict the binding affinity, several software or website-based tools have been developed now. These methods mainly fall into two categories: MHC-specific, which a method is applied to individual MHC molecule, such as NetMHC [22], SMM [23], SYFPEITHI [24]; and pan-specific, which a single method is applied to nearly all MHC-I molecules, such as NetMHCpan [25] and NetMHCcons [26]. In particular, the pan-specific method is very powerful tool, considering the huge polymorphism of MHC genes [27].…”

Peptide-Binding Groove Contraction Linked to the Lack of T Cell Response: Using Complex Structure and Energy To Identify Neoantigens

“…This conclusion is based on the fact that the K d values of most TCR ternary complexes are in the low-affinity range of ;0.1-500 mM (7). The six charged residues also hamper the interaction of 256 ILD with TCR because a hallmark of immunogenic CD8+ T cell epitopes is reportedly the possession of central, bulky, and hydrophobic (including aromatic) residues (51,52 , we retrospectively predicted T cell responses to the 12 oligopeptides using 11 publicly available residue-based methods (9)(10)(11)(12)(13)(14)(15)(16)(17)(18) and compared these predictions with those obtained with SE FF12MC using the robust Z score standardization protocol to minimize the influence of outliers (43). In this study, the immunogenicity predictions with the 11 residue-based methods used a reported assumption that both the affinity and stability of an oligopeptide × HLA complex are a correlate of immunogenicity (17).…”

“…One category comprises residue-based (or sequence-based) methods (such as those reported in Refs. [9][10][11][12][13][14][15][16][17][18]. These methods predict the HLA-complexation affinity or stability of a peptide with coarse granularity at the residue level.…”