Virus-Host Identification

This is a [nice paper](http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0027631) from my old adviser that attempts to identify what host a virus can infect based on its genome using multiclass Adaboost. It'd be cool to reproduce this with work with a DNN. Also the author is currently a postdoc at Stanford.

Furthermore, we could expand on this to find more subtle patterns. For example, this was a [really cool paper](http://mbio.asm.org/content/5/5/e01484-14.full) from UCSF that found ball pythons being infected with a virus of possible mouse-origin. If we could make a model that can predict the potential of these sorts of cross-species events (a.k.a zoonoses), it would be a game-changer for viral epidemiology, especially for diseases like influenza.

Start by reading the linked virus-host and zoonosis papers to identify the datasets, labels, and baseline methods needed for a bounded reproduction. No repository files, tests, or entry points are mentioned. Done would require a defined DNN experiment that reproduces the baseline and reports whether cross-species prediction is feasible.

Written by the indexing model from the issue text.