Amino Acid Encoding Deep Learning Applications

The repository contains the scripts and the models used for the paper amino acid encoding using deep learning application.

Environments and dependencies

we recommend creating a new conda environment using the config file defined at: resource/config.yml

Examples

All the models developed and mentioned in the paper can be retrained using the training script defined at CustomTrainingScripts directory.

train a DPPI model

$python trainDPPIModel.py -n 21 -d 8 -t 1 -g 0 -f 0.75 -o results/example_one

the above commandline would create a DPPI model using the blueprint defined at Models/DPPIBluePrint.py with a learned embedding of size 8, train it 75% of the training data on the first GPU on the system and write the results to results/example_one. To mark the embedding frozen the following command can be executed

$python trainDPPIModel.py -n 21 -d 8 -t 0 -g 0 -f 0.75 -o results/example_one

Funding

the project was funded by the German Research Foundation (DFG) (Research Training Group 1743, ‘Genes, Environment and Inflammation’) and by NASA Astrobiology Institute.