DeepHLApan is a deep learning approach used for predicting high-confidence neoantigens by considering both the presentation possibilities of mutant peptides and the potential immunogenicity of pMHC. The stand-alone software is available in github.

The architecture of DeepHLApan
437,077 HLA-pairs covering 81 HLA alleles were used for training the binding model.
32,785 HLA-peptide pairs with 7,212 of them being immunogenic were used for training immunogenicity model.