Adversarial autoencoders for MHCI epitope and immunogenicity prediction

Generative network-based approaches allow to use less training data, which is critical for MHCI with the lack of high-quality data.