CrossBind: Collaborative Cross-Modal Identification of Protein Nucleic-Acid-Binding Residues
Codes of protein residue binding prediction.
(1) Pytorch version.
conda install pytorch==1.4.0 cudatoolkit=10.1 -c pytorch
(2) Compile SparseConvNet ops.
cd lib/
python setup.py develop
(1) You can download the dataset (DNA/RNA pdb files) (http://www.csbio.sjtu.edu.cn/bioinf/GraphBind/ , https://github.com/biomed-AI/GraphSite)
(2) To transfer original PDB files to XYZ files, you need prepare LIG_TOOL.
git clone https://github.com/realbigws/PDB_Tool.git
(3) Modify the path in datasets/prepare_pdb_to_xyz.py
and run
cd datasets/
python prepare_pdb_to_xyz.py
(1) Load pre-trained structure representation DNA_127
/models/DNA_127_Structure.pkl
DNA_181
/models/Test_181.pkl
(2) Load ESM2 representation The details are described in (https://github.com/facebookresearch/esm)
Fine tune the CrossBind modal
Before the training phase, you can modify your model setting in cfgs/*yaml
, and choose the yaml
file you want to use.
Run full version of CrossBind:
python CrossBind.py --log_dir SparseConv_default --cfg_file cfgs/SparseConv-Cath-Decoys-Clf-Only.yaml --gpu 0
Visualization Case