Protein function prediction using a variational autoencoder. You can find the short paper which was an oral presentation at NIPS MLCB workshop 2017.
This repository contains the detailed workflow, and is associated with this blog post.
You can find more alignments/validation data to analyze from Marks' lab. Also, their group has done a comprehensive analysis on the application of VAE for all of this data independently.