KinasepKipred
Kinases to predict Inhibitor constant in terms of pKI (where pKi is decadic logarithm of Ki). We used the data points that were specifically represent Ki values to train and test the models.
Requirements:
- python==2.7.16
- pydpi==1.0
- numpy==1.16.5
- pandas==0.24.2
- tqdm==4.36.1
- scipy==1.2.2
- scikit-learn==0.20.4
- rdkit==2018.03
Setup
- Download the model file
./download_model.sh
- Set up the conda environment and activate it
conda env create -f environment.yml
conda activate kinasepki
Usage:
Run ./test.sh
to get the prediction for an example pair of protein sequence and a ligand SMILES. For any other inputs, run the following
python2 get_kinase_pki.py "<PROTEIN_SEQUENCE>" "<LIGAND_SMILES>"
Docker
- Build the docker image
docker build -t kinasepki .
and rundocker run --rm kinasepki
for a sample run that usestest.sh
. To provide sequence and SMILES, dodocker run --rm kinasepki <protein_sequence> <compound_smiles>
A docker image is also available on docker hub.
- Download the docker image
docker pull sirimullalab/kinasepkipred:py2
- Run the container
docker run --rm sirimullalab/kinasepkipred:py2 <protein_sequence> <compound_smiles>
. To run with a built-in sample, dodocker run --rm sirimullalab/kinasepkipred:py2