Table 1.
Methods and platforms commonly used for predicting neoantigens.
| Method [Ref] | Principle | Year |
|---|---|---|
| NetMHCpan [37] |
Comparison of epitope sequences by artificial neural networks that provide peptide–MHC-I-affinity predictions | 2016 |
| NetMHCIIpan [61] |
Pan-specific predictor able to predict binding affinities for all HLA-class-II molecules based on neural networks | 2013 |
| MHCflurry [62] |
Neutral networks including mass-spectrometry datasets for predicting peptide–MHC-I affinities | 2018 |
| ConvMHC [63] |
peptide–MHC interactions encoded into image-like array data and analyzed by deep convolutional neural network | 2017 |
| PLAtEAU [64] |
Defines shared consensus epitopes arising from a series of eluted nested peptides and quantified by mass spectrometry | 2018 |
| MuPeXI [65] |
Integration of somatic mutation calls, list of HLA types, an optional gene-expression profile, and NetMHCpan 3.0 to provide immunogenicity score based on similarity to non-mutated wild-type peptide | 2017 |
| NeoPrepPipe [66] |
Predicts neoantigen burdens and provide insights into the tumor heterogeneity, somatic mutation calls, and patient HLA haplotypes | 2019 |
| EpitopeHunter [67] |
Integrates expression of RNA with artificial neutral networks of immunogenicity-prediction algorithm based on the hydrophobicity of the TCR contact residues | 2015 |
| Neopepsee [68] |
Integrates sequence and amino-acid-immunogenicity information, including antigen processing and presentation to reduce the false-discovery rate | 2018 |