lylybig/MurJ_Inhibitor_Generation_dev

The Code for "MurJ Inhibitor Generation" Project

---

1. Installation

1.1 Dependency for main experiment

The code has been tested in the following environment:

Package	Version
Python	3.8.13
PyTorch	1.12.1
CUDA	11.0
PyTorch Geometric	2.5.2
RDKit	2021.03.1b1

Install via Conda and Pip:

conda create -n "sbdd" python=3.8.13
conda activate sbdd
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
pip install torch_geometric
pip install https://data.pyg.org/whl/torch-1.12.0%2Bcu113/pyg_lib-0.3.1%2Bpt112cu113-cp38-cp38-linux_x86_64.whl
pip install https://data.pyg.org/whl/torch-1.12.0%2Bcu113/torch_cluster-1.6.0%2Bpt112cu113-cp38-cp38-linux_x86_64.whl
pip install https://data.pyg.org/whl/torch-1.12.0%2Bcu113/torch_scatter-2.1.0%2Bpt112cu113-cp38-cp38-linux_x86_64.whl
pip install https://data.pyg.org/whl/torch-1.12.0%2Bcu113/torch_sparse-0.6.16%2Bpt112cu113-cp38-cp38-linux_x86_64.whl
pip install https://data.pyg.org/whl/torch-1.12.0%2Bcu113/torch_spline_conv-1.2.1%2Bpt112cu113-cp38-cp38-linux_x86_64.whl
conda install -c conda-forge "rdkit=2023.09.6"
conda install openbabel tensorboard pyyaml easydict python-lmdb -c conda-forge
pip install wandb
pip install pytorch-lightning==2.1.3
pip install biopython==1.83
pip install matplotlib
pip install numpy==1.23
pip install accelerate
pip install transformers
pip install gdown
pip install sympy==1.12
pip install networkx==2.8.5


# For Vina Docking
pip install meeko==0.1.dev3 scipy pdb2pqr vina==1.2.2
python -m pip install git+https://github.com/Valdes-Tresanco-MS/AutoDockTools_py3
conda install -c conda-forge autodock-vina -y

The code should work with PyTorch >= 1.9.0 and PyG >= 2.0. You can change the package version according to your need.

---

2. Universal Inference for a Specified Protein Pocket

Use sample_for_specific_protein.py to generate ligands for an arbitrary single protein pocket PDB.

2.1 Input Preparation

1. Prepare a pocket PDB centered at the binding site (e.g., 10 Å around thGoue ligand or binding residues).
   You may reuse the script in 4.1: ./data_preprocess/extract_pockets.py.

2. Example pocket file: ./specific_protein/3cl_ligand_pocket10.pdb.

3. The pretrained Benchmarked models are organized in the SBDD_Checkpoint Google Drive folder. Please download the checkpoints using the below script:

## NucleusDiff
gdown --fuzzy https://drive.google.com/file/d/1kA2ygcWKEcne0mMJM3lirmXKnhQnns14/view?usp=drive_link
## TargetDiff
gdown --fuzzy https://drive.google.com/file/d/1feT84RrmQcRJPVFsUfq7ZJHX_vIZWxdl/view?usp=drive_link

, and save it to ./checkpoints.

2.2 Inference (sampling)

python sample_for_pocket.py \
        --model_type nucleusdiff \
        --checkpoint ./checkpoints/nucleusdiff_pretrained_model.pt \
        --pdb_path ./specific_protein/3cl_ligand_pocket10.pdb \
        --result_path ./results_specific_protein \
        --sample_num_atoms real_world_testing \
        --inference_num_atoms 30 \
        --num_samples 1000 \
        --num_steps 1000 \
        --device cuda:0

Key arguments:

• --model_type: the model we use to do inference, [targetdiff, nucleusdiff].
• --checkpoint: path to a NucleusDiff checkpoint (.pt).
• --pdb_path: pocket PDB for your target protein.
• --result_path: output directory.
• --sample_num_atoms: set to real_world_testing to use a fixed atom count.
• --inference_num_atoms: atoms per generated ligand when using real_world_testing.
• --num_samples: number of ligands to generate.
• --num_steps: diffusion steps (trade-off between quality and speed).
• --device: GPU device, e.g., cuda:0.

2.3 Outputs

• ${result_path}/sample_{test_time}.pt: raw tensors and sampling trajectories.
• ${result_path}/sdf/*.sdf: reconstructed molecules in SDF format.

Run python sample_for_specific_protein.py --help for the complete list of options and defaults.

---

3. Binding Affinity Evaluation of Given Peptide

Note: For peptides (with many rotatable bonds), using Vina via the command-line interface is more robust and reproducible than through a Python wrapper because:

• You explicitly control preparation and parameters (box, exhaustiveness, seed).
• You avoid hidden defaults in wrappers that peptides are sensitive to (conformer generation, box sizing).
• Dependencies and errors are more transparent, making debugging easier.
• It’s easier to tweak individual steps (protonation, box size, conformers) as needed.

3.1 Dependencies (required; already satisfied if you're using our recommended environment)

vina, obabel, pdb2pqr30, AutoDockTools

3.2 Bash Scripts

bash ./scripts/step_04_run_vina_chainB.sh

---

4. Evaluation for DiffBP

4.1 Training data

gdown --fuzzy https://drive.google.com/file/d/1HGzwpqZGoBaosPGEnn3Lb_b-iNLTbGVO/view?usp=drive_link
gdown --fuzzy https://drive.google.com/file/d/1bFR5IfnV5HGAF58XjXfn-xiVKoV83kfB/view?usp=drive_link
gdown --fuzzy https://drive.google.com/file/d/1KD3MGbcsJ2Q5Cn7rZ7ghkOAr-pxgyarm/view?usp=drive_link

4.2 Inference