conda install mamba
mamba env create -f surfgen_env.yml -n surfgen
conda activate surfgen
If you're reluctant to use mamba:
conda env create -f surfgen_env.yml -n surfgen
We also provide conda-packed file here. Download it and then unzip it in your conda/envs/dir. For me, the directory is ~/.conda/envs. Special thanks to the creators and organizers of zenodo, which provides a free platform to store large files for academic use.
mkdir ~/.conda/envs/surfgen
tar -xzvf surfgen.tar.gz -C ~/.conda/envs/surfgen
conda activate surfgen
Since Nvidia 40 series cards no longer support CUDA 11.3, I also created the SurfGen environment for an RTX-4080 powered linux system.
mamba create -n surfgen pytorch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 pytorch-cuda=12.1 plyfile pyg rdkit biopython easydict jupyter ipykernel lmdb -c pytorch -c nvidia -c pyg -c conda-forge
pip install lmdb
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.2.0+cu121.html
Note: PyG subgraph function has been put to another place, therefore, replace the following command at ./utils/transform.py
# from torch_geometric.utils.subgraph import subgraph
from torch_geometric.data.data import subgraph
The main data used for training is CrossDock2020
wget https://bits.csb.pitt.edu/files/crossdock2020/CrossDocked2020_v1.1.tgz -P data/crossdock2020/
tar -C data/crossdock2020/ -xzf data/crossdock2020/CrossDocked2020_v1.1.tgz
wget https://bits.csb.pitt.edu/files/it2_tt_0_lowrmsd_mols_train0_fixed.types -P data/crossdock2020/
wget https://bits.csb.pitt.edu/files/it2_tt_0_lowrmsd_mols_test0_fixed.types -P data/crossdock2020/
Then follow the guidelines to process it. The train data split is split_name.pt.
If it's inconvenient for you, we also provided the processed data. You just need to download them in ./data and create a ./data/crossdock_pocket10 directory, and put the index.pkl in it.
Although we have prepared the required data for training and evaluation above. But you may want to apply SurfGen in your own case. So we provide the guidelines for creating the surf_maker environment.
conda create -n surf_maker pymesh2 jupyter scipy joblib biopython rdkit plyfile -c conda-forge
We highly recommend using mamba instead of conda for speeding up.
mamba create -n surf_maker pymesh2 jupyter scipy joblib biopython rdkit plyfile -c conda-forge
We also provide the .yml file for creating environment
conda env create -f surf_maker_environment.yml
When the base python environment was created, then install APBS-3.0.0, pdb2pqr-2.1.1 on your computer. Then set the msms_bin, apbs_bin, pdb2pqr_bin, and multivalue_bin path directly in your ~/.bashrc, or just set them in the scripts when creating the surface file from the pdb file.
Now you have deployed all the dependent environments. Please follow the ./data/surf_maker for making surface data. Or run the ./data/surf_maker/surf_maker_test.py for testing whether you have figured out this environment successfully.
python ./data/surf_maker/generate_surface.ipynb
If the surface is generated, you will find the .ply file in the ./data/surf_maker
And we provide the generated surface file at ./data, namely 3cl_pocket_8.0_res_1.5.ply for further generation.
To generate the example, run the gen.py. The model's parameters can be downloaded here. Put it at ./ckpt.
We provide an example of the pharmaceutic target for Covid-19, 3cl protein, in the ./example, run the following code to generate inhibitors directly inside the pocket!
python gen.py --outdir example --check_point ./ckpt/val_119.pt --ply_file ./example/3cl_pocket_8.0_res_1.5.ply
python train.py
For surface generation, a common error is:
No such file or directory: '/tmp/tmpc5aa wvj/temp1_out.csv'
This error primarily originates from APBS tools. Breaking down the code reveals the exact problem:
error while loading shared libraries: libTABIPBlib.so: cannot open shared object file: No such file or directory
This occurs because the APBS library is not included in the LD_LIBRARY_PATH
For Ubuntu 18 system, once you download APBS-3.0.0 (~300MB) and pdb2pqr-2.1.1 on your computer, like:
Then, add the LD_LIBRARY_PATH
to your ~/.bashrc
, for example:
# Install Vim if necessary:
sudo apt install vim
# Edit ~/.bashrc
vim ~/.bashrc
# Append the following command at the end of the ~/.bashrc:
...
export LD_LIBRARY_PATH="/home/haotian/software/miniconda3/envs/deepdock/lib:$LD_LIBRARY_PATH"
...
# Save and exit Vim, then activate the setting:
source ~/.bashrc # active the setting
However, sometimes another error might occur:
libreadline.so.7: cannot open shared object file: No such file or directory
When I encountered this problem, I was using Ubuntu 22. I found that libreadline.so.7
is only available for Ubuntu 18, and there is no easy way to install libreadline.so.7
on Ubuntu 22 system. Eventually, I found a solution by downloading both APBS-3.0.0
and APBS-3.4.1
. I stored these two software in Zenodo.
Assign the paths at ./utils/masif/generate_prot_ply.py
as follows:
msms_bin="{install_path}/APBS-3.0.0.Linux/bin/msms"
apbs_bin = '{install_path}/APBS-3.4.1.Linux/bin/apbs'
pdb2pqr_bin="{install_path}/pdb2pqr-linux-bin64-2.1.1/pdb2pqr"
multivalue_bin="{install_path}/APBS-3.0.0.Linux/share/apbs/tools/bin/multivalue"