Protein patch variational autoencoder (VAE) to characterize protein surface topography & chemistry
This project is designed for the creation of latent salient features for inhomogeneous protein surface patches using a 3D convolutional variational autoencoder. The project has three components:
- Creation of patch voxels from raw protein PDB files
- Training of a 3D Convolutional VAE using patch voxel dataset, and subsequent creation of patch latent vectors
- Training of a property prediction 3D Conv VAE for patch dataset
Please contact us (sinhaimee@gmail.com) before using any of this code.
