DeepContact

Installation

1. Install the following dependencies

• HHBlits 2.0.16
• Jackhmmer 3.1b2
• HHFilter 2.0.16
• CCMPred
• FreeContact 1.0.21
• MetaPsicov 1.02
• Blast 2.2.6
• Psipred 4.0

2. We need set environment variable export BLASTDB=/data/work/yang/Qing/databases/data/nr
3. Modify default.yaml with path to the above programs
4. Install python dependencies with

conda env create -f environment.yml
source activate deepcontact-env
pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip

Usage Example

Predict using all features

python data-processing/run_pipeline.py default.yaml test.fasta ./tmp_feature
python deepcontact/feature_gen.py ./deepcontact/feature.yaml ./tmp_feature ./tmp_pickle/feature.pkl
python deepcontact/main.py ./tmp_pickle/feature.pkl ./tmp_output/prediction.pkl

Predict using only CCMPred Predictions

• feature_gen parameters:
  • a config including the feature names.
  • the feature directory ( this dir of your ccmpred outputs)
  • the pickled output (default will be tmp_pickle)

python deepcontact/feature_gen.py ./deepcontact/feature_only_ccmpred.yaml ./tmp_feature ./tmp_pickle/feature_only_ccmpred.pkl
python deepcontact/main_only_ccmpred.py ./tmp_pickle/feature_only_ccmpred.pkl ./tmp_output/prediction_only_ccmpred.pkl