kailiwang1 / deepdtaf

a deep learning architecture for protein-ligand binding affinity prediction

About DeepDTAF

DeepDTAF is a deep learning architecture, which integrates local and global features to predict the binding affinity between ligands and proteins.

The benchmark dataset can be found in ./data/. Data preprocessing can be referred to ./prepare/. The DeepDTAF model is available in ./src/. And the result will be generated in ./runs/. See our paper for more details.

Requirements:

• python 3.7
• cudatoolkit 10.1.243
• cudnn 7.6.0
• pytorch 1.4.0
• numpy 1.16.4
• scikit-learn 0.21.2
• pandas 0.24.2
• tensorboard 2.0.0
• scipy 1.3.0
• numba 0.44.1
• tqdm 4.32.1

The easiest way to install the required packages is to create environment with GPU-enabled version:

conda env create -f environment_gpu.yml
conda activate DeepDTAF_env

Then, install the apex in the DeepDTAF_env environment:

git clone https://github.com/NVIDIA/apex
cd apex
python setup.py install

Since the codes for apex package in the above website have been updated, you can also install it using the uploaded package. The apex.tar can be found in ./src/.

Training & Evaluation

to train your own model

cd ./src/
python main.py

to see the result

tensorboard ../runs/DeepDTAF_<datetime>_<seed>/

Citation

Wang K, Zhou R, Li Y, et al. DeepDTAF: a deep learning method to predict protein–ligand binding affinity. Briefings in Bioinformatics, 2021, 22(5), bbab072.

contact

Kaili Wang: [email protected]

deepdtaf's Issues

Question About Datasets

Hi, I have a question.

DeepDTAF paper uses three test sets, but I have only found two, and one 71 data (test71) has not been found. I urgently need this data set for testing. Finally, if I can send it to me
This is my email:[email protected]

Could yor correct the link of table supplements please?

Question about physiochemical feature

Hi, I have a question.

DeepDTAF paper proposes some physiochemical features but all link of table supplements of the paper is not proper. As a result, I cannot understand physiochemical features used in the paper.

Could yor correct the link of table supplements please?

Can you provide code for PDBbind raw data processing to csv？