This repository describe the usage of neural networks to predict the affinity of peptides to MHC type I for humans. Neural networks was written in tensorflow and keras.
MHC class I affinity is usually used as a proxy for the description of a good epitopes. Given a dataset, (some of which can be found http://tools.immuneepitope.org/main/datasets/) it is possible to learn how to predict the affinity of a given epitode to an MHC class I receptor.
This approach can work for data of any MHC class I or II, and for any species, given that sufficient data are provided.