Skip to content
This repository was archived by the owner on Sep 26, 2023. It is now read-only.
/ PUResNet Public archive
forked from jivankandel/PUResNet

Predicting protein-ligand binding sites using deep convolutional neural network

Notifications You must be signed in to change notification settings

nuriafari/PUResNet

 
 

Repository files navigation

PUResNet(Predicting protein-ligand binding sites using deep convolutional neural network)

Prediction of protein-ligand binding site is fundamental step in understanding functional characteristics of the protein which plays vital role in carrying out different biological functions and is a crucial stage in drug discovery. A protein shows its true nature after interacting with any capable molecule knows as ligand which binds only in favorable binding site of protein structure.

Training Dataset

scpdb_subset.zip (https://github.com/jivankandel/PUResNet/blob/main/scpdb_subset.zip)

Independent Dataset

coach.zip (https://github.com/jivankandel/PUResNet/blob/main/coach.zip)
BU48.zip (https://github.com/jivankandel/PUResNet/blob/main/BU48.zip)

Requirements

  1. Tensorflow 1.11 (https://www.tensorflow.org/)
  2. Keras (https://keras.io/)
  3. Scipy (https://www.scipy.org/)
  4. Scikit-Image (https://scikit-image.org/)
  5. Open Babel (http://openbabel.org/wiki/Main_Page)
  6. Pybel (http://openbabel.org/docs/current/UseTheLibrary/Python_Pybel.html)
  7. TFBIO (https://gitlab.com/cheminfIBB/tfbio)
  8. Numpy (https://numpy.org/)
  9. Python 3.6 (https://www.python.org/)
    Note that: It is better to setup new environment using conda or pyenv. You may need to compile open babel and tfbio if installing with PIP doesn't work.

Model Architecture

Figure showing Convolutional Block,Identiy Block and Up Sampling Block

Figure showing Architecture of PUResNet

Usage

1. Clone this repository
git clone https://github.com/jivankandel/PUResNet.git
cd PUResNet
2. Setup Environment
#create conda environment
conda create -n env_name python=3.6 
conda activate env_name
conda install -c conda-forge openbabel
conda install scikit-image
conda install numpy
conda install -c anaconda scipy
conda install -c conda-forge keras=2.1
conda install -c conda-forge tensorflow=1.11
2. Prediction
python predict.py -h
usage: predict.py [-h] --file_format FILE_FORMAT --mode MODE --input_path
                  INPUT_PATH --model_path MODEL_PATH
                  [--output_format OUTPUT_FORMAT] [--output_path OUTPUT_PATH]
                  [--gpu GPU]

optional arguments: -h, --help show this help message and exit --file_format FILE_FORMAT, -ftype FILE_FORMAT File Format of Protein Structure like: mol2,pdb..etc. All file format supported by Open Babel is supported (default: None) --mode MODE, -m MODE Mode 0 is for single protein structure. Mode 1 is for multiple protein structure (default: None) --input_path INPUT_PATH, -i INPUT_PATH For mode 0 provide absolute or relative path for protein structure. For mode 1 provide absolute or relative path for folder containing protein structure (default: None) --model_path MODEL_PATH, -mpath MODEL_PATH Provide models absolute or relative path of model (default: None) --output_format OUTPUT_FORMAT, -otype OUTPUT_FORMAT Provide the output format for predicted binding side. All formats supported by Open Babel (default: mol2) --output_path OUTPUT_PATH, -o OUTPUT_PATH path to model output (default: output) --gpu GPU, -gpu GPU Provide GPU device if you want to use GPU like: 0 or 1 or 2 etc. (default: None)

Example:

python predict.py -ftype pdb -m 0 -i test.pdb -mpath whole_trained_model1.hdf -otype mol2 -o output -gpu 1

About

Predicting protein-ligand binding sites using deep convolutional neural network

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 96.7%
  • Jupyter Notebook 3.3%