On generative models of T-cell receptor sequences

In this comment on Davidsen et al., "Deep generative models for T cell receptor protein sequences", eLife 2019;8:e46935, we compare the performance of the variational auto-encoder presented in that article to a previously proposed approach for which a software implementation, SONIA, has been recently released. We find that SONIA performs as well as the variational auto-encoder, at a lower computational cost.