Code for the paper "MUST-CNN: A Multilayer Shift-and-Stitch Deep Convolutional Architecture for Sequence-Based Protein Structure Prediction" (AAAI 2016)
Zeming Lin, Jack Lanchantin, Yanjun Qi
University of Virginia
Code is written in Lua and requires [Torch] (http://torch.ch/). Running on the GPU via the cuda cunn library is strongly reccommended (use the -cuda parameter). To get cunn, use:
luarocks install cunn
Att: you might need to run luarocks install class on some computers.
The data is split up into 2 directories: 4Protein, and cb513. Each directory contains a "data" subdirectory and a "hash" subdirectory. The data subdirectory contains "aa1.dat" which is the raw protein sequence data, as well as each *tag.dat file which are the class labels for each separate class. The data subdir also contains the psi-blast files. The hash subdirectory contains the dictionary numbers for each of the amino acids and class labels.
The data directories are included in this repository as tar files. Untar the data directory which you choose to use with the -dataset parameter (4Protein is the default dataset).
tar -xvf ./data/4Protein.tar.gz -C ./data/
The code should be runnable with the default parameters by simply executing:
th main.lua
See [cmdlineargs.lua] (https://github.com/DeepLearning4BioSeqText/Paper16-AAAI-MUST-CNN/blob/master/cmdlineargs.lua) to pass in parameters.
e.g.,
th main.lua -model mlp
th main.lua -cuda