DiffPepBuilder

https://github.com/yuzhewangpku/diffpepbuilder
Admin
DiffPepBuilder is a diffusion-based framework for de novo peptide binder design, protein-peptide...
Tokens:6,332
Snippets:59
Trust Score:7.7
License:MIT
Update:1 month ago
Show doc for...
Tokens:
Raw
### Install PyRosetta

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Install the PyRosetta wheel required for post-processing tasks.

```bash
wget https://west.rosettacommons.org/pyrosetta/release/release/PyRosetta4.MinSizeRel.python39.linux.wheel/pyrosetta-2024.39+release.59628fb-cp39-cp39-linux_x86_64.whl
pip install pyrosetta-2024.39+release.59628fb-cp39-cp39-linux_x86_64.whl
```

--------------------------------

### Install DiffPepDock dependencies

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Installs required Python packages and the pdbfixer tool.

```python
#@title ### Install dependencies
os.system("pip install wget wandb fair-esm biotite pyrootutils easydict biopython tqdm ml-collections mdtraj GPUtil dm-tree tmtools py3Dmol")
os.system("wget https://west.rosettacommons.org/pyrosetta/release/release/PyRosetta4.MinSizeRel.python312.linux.wheel/pyrosetta-2024.39+release.59628fb-cp312-cp312-linux_x86_64.whl")
os.system("pip install pyrosetta-2024.39+release.59628fb-cp312-cp312-linux_x86_64.whl")
os.system("rm pyrosetta-2024.39+release.59628fb-cp312-cp312-linux_x86_64.whl")

pdbfixer_folder = "pdbfixer"
if not os.path.isdir(pdbfixer_folder):
  print("Installing pdbfixer...")
  os.system("git clone https://github.com/openmm/pdbfixer.git")
  os.chdir(pdbfixer_folder)
  os.system("python setup.py install")
  os.chdir("..")
  print("pdbfixer is installed.")
else:
  print("pdbfixer is already cloned.")

os.system("pip install hydra-core hydra-joblib-launcher")
```

--------------------------------

### Create Conda Environment

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Set up the required dependencies using the provided environment file.

```bash
conda env create -f environment.yml
conda activate diffpepbuilder
```

--------------------------------

### Download DiffPepDock model assets

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Downloads the model checkpoint and sets up the directory structure if not already present.

```python
#@title ### Download model assets
import os

diffpep_folder = "DiffPepBuilder"
checkpoint_file = os.path.join(diffpep_folder, "experiments", "checkpoints", "diffpepdock_v1.pth")

if not (os.path.isdir(diffpep_folder) and os.path.isfile(checkpoint_file)):
  print("Installing DiffPepDock...")

  if not os.path.isdir(diffpep_folder):
    os.system("git clone https://github.com/YuzheWangPKU/DiffPepBuilder.git")

  os.chdir(diffpep_folder)

  if not os.path.isfile(checkpoint_file):
    print("Downloading model weights...")
    os.system("wget https://zenodo.org/records/15398020/files/diffpepdock_v1.pth")
    os.makedirs("experiments/checkpoints/", exist_ok=True)
    os.system("mv diffpepdock_v1.pth experiments/checkpoints/")

  os.chdir("..")
  print("DiffPepDock is installed and ready.")

else:
  print("DiffPepDock is already installed and ready.")
```

--------------------------------

### Finetune Model for Docking

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Runs the finetuning process for docking tasks using the specified configuration file.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
torchrun --nproc-per-node=8 experiments/train.py --config-name=finetune
```

--------------------------------

### Download Model Weights

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Downloads the required model weights from Zenodo and moves them to the appropriate directory.

```bash
wget https://zenodo.org/records/15398020/files/diffpepdock_v1.pth
mv diffpepdock_v1.pth experiments/checkpoints/
```

--------------------------------

### Download Model Weights

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Downloads the required model weights from Zenodo and moves them to the appropriate directory.

```bash
wget https://zenodo.org/records/12794439/files/diffpepbuilder_v1.pth
mv diffpepbuilder_v1.pth experiments/checkpoints/
```

--------------------------------

### Download and Extract Training Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Downloads the raw dataset from Zenodo and extracts it into the local data directory.

```bash
wget https://zenodo.org/records/13744959/files/PepPC-F_raw_data.tar.gz
mkdir data/PepPC-F_raw_data
tar -xvf PepPC-F_raw_data.tar.gz --strip-components=1 -C data/PepPC-F_raw_data
```

--------------------------------

### Run Post-processing

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Performs side chain reconstruction and optimization on the generated peptide binders.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
python experiments/run_postprocess.py --in_pdbs runs/inference --ori_pdbs examples/receptor_data --amber_relax --rosetta_relax
```

--------------------------------

### Download DiffPepBuilder model assets

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Downloads the repository, model weights, and required libraries if they are not already present in the local environment.

```python
#@title ### Download model assets
import os

diffpep_folder = "DiffPepBuilder"
ssbuilder_lib = os.path.join(diffpep_folder, "SSbuilder", "SSBLIB")
checkpoint_file = os.path.join(diffpep_folder, "experiments", "checkpoints", "diffpepbuilder_v1.pth")

if not (os.path.isdir(diffpep_folder) and os.path.isdir(ssbuilder_lib) and os.path.isfile(checkpoint_file)):
  print("Installing DiffPepBuilder...")

  if not os.path.isdir(diffpep_folder):
    os.system("git clone https://github.com/YuzheWangPKU/DiffPepBuilder.git")

  os.chdir(diffpep_folder)

  if not os.path.isdir(ssbuilder_lib):
    print("Installing SSBLIB...")
    os.system("tar -xvf SSbuilder/SSBLIB.tar.gz -C SSbuilder")

  if not os.path.isfile(checkpoint_file):
    print("Downloading model weights...")
    os.system("wget https://zenodo.org/records/12794439/files/diffpepbuilder_v1.pth")
    os.makedirs("experiments/checkpoints/", exist_ok=True)
    os.system("mv diffpepbuilder_v1.pth experiments/checkpoints/")

  os.chdir("..")
  print("DiffPepBuilder is installed and ready.")

else:
  print("DiffPepBuilder is already installed and ready.")
```

--------------------------------

### Download PepPC raw data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/datasets/README.md

Use wget to download the compressed raw data archive for the PepPC dataset from Zenodo.

```bash
wget https://zenodo.org/records/13744959/files/PepPC_raw_data.tar.gz
```

--------------------------------

### Download PepPC-F raw data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/datasets/README.md

Use wget to download the compressed raw data archive for the PepPC-F dataset from Zenodo.

```bash
wget https://zenodo.org/records/13744959/files/PepPC-F_raw_data.tar.gz
```

--------------------------------

### Clone Repository

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Download the source code from the GitHub repository.

```bash
git clone https://github.com/YuzheWangPKU/DiffPepBuilder.git
cd DiffPepBuilder
```

--------------------------------

### Train DiffPepBuilder Model

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Initiates the training process using torchrun. Ensure the BASE_PATH environment variable is set correctly.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
torchrun --nproc-per-node=8 experiments/train.py
```

--------------------------------

### Execute Protein-Peptide Docking

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Runs the docking process using the preprocessed metadata. Ensure the BASE_PATH is set correctly and adjust nproc-per-node based on available GPU resources.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
torchrun --nproc-per-node=8 experiments/run_docking.py data.val_csv_path=data/docking_data/metadata_test.csv
```

--------------------------------

### Execute Protein-Peptide Docking

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Configures the base path and runs the docking script using torchrun. Ensure the environment variable BASE_PATH is set correctly before execution.

```python
#@title ### Run docking
#@markdown - Postprocessing will be performed automatically after docking.
os.environ['BASE_PATH'] = "/content/DiffPepBuilder"

!torchrun --nproc-per-node=1 DiffPepBuilder/experiments/run_docking.py \
  data.val_csv_path=test_case/metadata_test.csv \
  experiment.use_ddp=False \
  experiment.num_gpus=1 \
  experiment.num_loader_workers=1
```

--------------------------------

### Download Docking Results

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Compresses the docking results directory into a tarball and triggers a browser download.

```python
#@title ### Download results
#@markdown - The docking results will be saved in the `DiffPepBuilder/runs` folder.

!tar --directory=/content/DiffPepBuilder/runs -czf /content/docking_results.tar.gz docking
files.download("/content/docking_results.tar.gz")
```

--------------------------------

### Customize Docking and Post-processing Settings

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Updates the docking configuration YAML file with sampling parameters and relaxation settings.

```python
#@title ### Customize docking settings
import yaml

#@markdown #### Sampling params
denoising_steps = "200" #@param [100, 200, 500]
noise_scale = "1" #@param [0.5, 1, 1.5, 2, 2.5]
samples_per_sequence = 4 #@param {type:"integer"}

#@markdown #### Postprocessing params
amber_relax = True #@param {type:"boolean"}
#@markdown  - Amber relaxation runs slowly on Colab. You may skip this step for faster inference with minimal quality loss.
rosetta_relax = True #@param {type:"boolean"}

yaml_file_path = "DiffPepBuilder/config/docking.yaml"
with open(yaml_file_path, 'r') as file:
  yaml_data = yaml.safe_load(file)

yaml_data['data']['num_t'] = int(denoising_steps)
yaml_data['experiment']['noise_scale'] = float(noise_scale)
yaml_data['experiment']['eval_batch_size'] = int(samples_per_sequence)
yaml_data['data']['num_repeat_per_eval_sample'] = int(samples_per_sequence)

yaml_data['postprocess']['amber_relax'] = amber_relax
yaml_data['postprocess']['rosetta_relax'] = rosetta_relax

with open(yaml_file_path, 'w') as file:
  yaml.dump(yaml_data, file, default_flow_style=False)
```

--------------------------------

### Run Inference

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Executes the inference script to generate peptide binders, requiring the base path to be set and GPU configuration.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
torchrun --nproc-per-node=8 experiments/run_inference.py data.val_csv_path=data/receptor_data/metadata_test.csv
```

--------------------------------

### Run Postprocessing

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Executes postprocessing on docking results to compute binding ddG values, optionally applying Amber and Rosetta relaxation.

```bash
export BASE_PATH="your/path/to/DiffPepBuilder"
python experiments/run_postprocess.py --in_pdbs runs/docking --ori_pdbs examples/docking_data --amber_relax --rosetta_relax
```

--------------------------------

### Run de novo generation

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Executes the inference script using torchrun to generate peptide binders based on the provided configuration.

```python
#@title ### Run *de novo* generation
os.environ['BASE_PATH'] = "/content/DiffPepBuilder"

!torchrun --nproc-per-node=1 DiffPepBuilder/experiments/run_inference.py \
  data.val_csv_path=test_case/metadata_test.csv \
  experiment.use_ddp=False \
  experiment.num_gpus=1 \
  experiment.num_loader_workers=1
```

--------------------------------

### Preprocess Docking Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Prepares target and peptide sequences for docking by processing input files into the specified directory.

```bash
python experiments/process_batch_dock.py --pdb_dir examples/docking_data --write_dir data/docking_data --receptor_info_path examples/docking_data/docking_cases.json --peptide_seq_path examples/docking_data/peptide_seq.fasta
```

--------------------------------

### Preprocess Training Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Processes raw PDB files into a structured dataset format suitable for training.

```bash
python experiments/process_dataset.py --pdb_dir data/PepPC-F_raw_data --write_dir data/complex_dataset
```

--------------------------------

### Configure Docking Mode and Sequence Input

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Uses ipywidgets to toggle between single and batch docking modes and save peptide sequences to a FASTA file.

```python
#@title ### Specify peptide sequences
import ipywidgets
from IPython.display import display, clear_output

mode_dropdown = ipywidgets.Dropdown(
    options=['single', 'batch'],
    value='single',
    description='Docking Mode:',
    style={'description_width': 'initial'}
)

# Widgets for SINGLE mode
peptide_seq = ipywidgets.Text(
    value='VLGEPRYAFNFN',
    description='Peptide Sequence:',
    style={'description_width': 'initial'}
)
peptide_id = ipywidgets.Text(
    value='nat',
    description='Peptide ID:',
    style={'description_width': 'initial'}
)
save_button = ipywidgets.Button(
    description='Save Sequence',
    button_style='success'
)

# Widgets for BATCH mode
upload_widget = ipywidgets.FileUpload(
    accept='.fasta',
    multiple=False,
    description='Upload FASTA'
)
process_button = ipywidgets.Button(
    description='Process Upload',
    button_style='info'
)

output = ipywidgets.Output()

def on_mode_change(change):
    with output:
        clear_output(wait=True)
        if change['new'] == 'single':
            display(ipywidgets.VBox([peptide_seq, peptide_id, save_button]))
        else:
            display(ipywidgets.VBox([upload_widget, process_button]))

def save_sequence(btn):
    seq = peptide_seq.value.strip()
    pid = peptide_id.value.strip()
    if not seq or not pid:
        with output:
            print("Please provide both Sequence and ID.")
        return
    path = os.path.join("test_case", "peptide_seq.fasta")
    with open(path, 'w') as f:
        f.write(f">{pid}\n{seq}\n")
    with output:
        print(f"Saved single-sequence file to {path}")

def process_upload(btn):
    if not upload_widget.value:
        with output:
            clear_output(wait=True)
            print("No FASTA file uploaded.")
        return
    with output:
        clear_output(wait=True)
        for fname, fileinfo in upload_widget.value.items():
            dest = os.path.join("test_case", "peptide_seq.fasta")
            with open(dest, 'wb') as f:
                f.write(fileinfo['content'])
            print(f"Saved batch file to {dest}")

mode_dropdown.observe(on_mode_change, names='value')
save_button.on_click(save_sequence)
process_button.on_click(process_upload)

display(mode_dropdown, output)
on_mode_change({'new': mode_dropdown.value})
```

--------------------------------

### Download generated results

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Compresses the inference directory into a tarball and triggers a browser download in a Jupyter environment.

```python
#@title ### Download results

!tar --directory=/content/DiffPepBuilder/runs -czf /content/designed_binders.tar.gz inference
files.download("/content/designed_binders.tar.gz")
```

--------------------------------

### Run Postprocessing Protocol

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Executes the postprocessing script on generated PDB files. This process is computationally intensive and recommended for local machines with high CPU availability.

```python
#@title ### (Optional) Run postprocessing
#@markdown - This step executes slowly on Colab; we recommend running it on a local machine with more CPU cores available
os.environ['BASE_PATH'] = "/content/DiffPepBuilder"

!python DiffPepBuilder/experiments/run_postprocess.py --in_pdbs /content/DiffPepBuilder/runs/inference --ori_pdbs /content/DiffPepBuilder/examples/receptor_data --nproc 2 --amber_relax --rosetta_relax
```

--------------------------------

### Customize generation settings

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Updates the inference configuration file with sampling parameters, peptide length constraints, and disulfide bond settings.

```python
#@title ### Customize generation settings
import yaml

#@markdown #### Sampling params
denoising_steps = "200" #@param [100, 200, 500]
noise_scale = "1" #@param [0.5, 1, 1.5, 2, 2.5]
seq_temperature = "0.1" #@param [0.1, 0.5, 1, 2.5, 10]

#@markdown #### Peptide length
min_length = 12 #@param {type:"integer"}
max_length = 16 #@param {type:"integer"}
samples_per_length = 4 #@param {type:"integer"}

#@markdown #### Disulfide bond settings
build_ss_bond = True #@param {type:"boolean"}
max_ss_bond = 2 #@param {type:"integer"}
entropy_threshold = 0.01 #@param [0.001, 0.01, 0.05, 0.1, 0.5]

yaml_file_path = "DiffPepBuilder/config/inference.yaml"
with open(yaml_file_path, 'r') as file:
  yaml_data = yaml.safe_load(file)

yaml_data['inference']['denoising']['num_t'] = int(denoising_steps)
yaml_data['inference']['denoising']['noise_scale'] = float(noise_scale)
yaml_data['inference']['sampling']['samples_per_length'] = samples_per_length
yaml_data['inference']['sampling']['min_length'] = min_length
yaml_data['inference']['sampling']['max_length'] = max_length
yaml_data['inference']['sampling']['seq_temperature'] = float(seq_temperature)
yaml_data['inference']['ss_bond']['build_ss_bond'] = build_ss_bond
yaml_data['inference']['ss_bond']['max_ss_bond'] = max_ss_bond
yaml_data['inference']['ss_bond']['entropy_threshold'] = float(entropy_threshold)

with open(yaml_file_path, 'w') as file:
  yaml.dump(yaml_data, file, default_flow_style=False)
```

--------------------------------

### Specify receptor information

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Configures the receptor PDB file and binding parameters for the docking task.

```python
#@title ### Specify receptor information
from google.colab import files
import json

os.makedirs("test_case", exist_ok=True)
receptor_type = "default (7Z6F)" #@param ["default (7Z6F)", "uploaded"]

if receptor_type == "uploaded":
  uploaded_pdb = files.upload()
  file_name = next(iter(uploaded_pdb))
  if not file_name.endswith(".pdb"):
    raise ValueError("Please upload in PDB format.")
  os.system(f"cp {file_name} test_case/")
  os.system(f"cp test_case/{file_name} DiffPepBuilder/examples/docking_data/")
else:
  file_name = "7Z6F.pdb"
  os.system(f"cp DiffPepBuilder/examples/docking_data/7Z6F.pdb test_case/")
#@markdown - **Note**: please remove non-protein components from the PDB file and ensure that the CA atoms are present.
#@markdown - **Note**: please remove any underscores (`_`) from the receptor PDB file name, as the script interprets `_` as a delimiter.

lig_chain = "A" #@param {type:"string"}
#@markdown  - Chain ID of the **reference** ligand. Please set to `None` if no reference ligand is included in the PDB file.
#@markdown  The model will prioritize reference ligand information over the binding motif if both are given.
motif = None #@param {type:"string"}
#@markdown  - Comma-separated residue IDs of the binding motif on the receptor. An example is `"B40, B58-59, B71-72, B87"`

key = os.path.splitext(file_name)[0]
data = {}
if lig_chain and lig_chain != "None":
  data['lig_chain'] = lig_chain
if motif and motif != "None":
  data['motif'] = motif.replace(",", "-")

json_file_write_path = "test_case/docking_cases.json"
final_data = {key: data}
with open(json_file_write_path, 'w') as file:
  json.dump(final_data, file, indent=4)
```

--------------------------------

### Preprocess receptor data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Runs the receptor processing script using the configuration generated in the previous step.

```python
#@title ### Preprocess receptors
!python DiffPepBuilder/experiments/process_receptor.py \
  --pdb_dir test_case \
  --write_dir test_case \
  --receptor_info_path test_case/de_novo_cases.json
```

--------------------------------

### Specify receptor information

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepBuilder_demo.ipynb

Configures the target receptor, ligand chains, and binding hotspots or motifs, saving the configuration to a JSON file.

```python
#@title ### Specify receptor information
from google.colab import files
import json

os.makedirs("test_case", exist_ok=True)
receptor_type = "default (ALK1)" #@param ["default (ALK1)", "uploaded"]

if receptor_type == "uploaded":
  uploaded_pdb = files.upload()
  file_name = next(iter(uploaded_pdb))
  if not file_name.endswith(".pdb"):
    raise ValueError("Please upload in PDB format.")
  os.system(f"cp {file_name} test_case/")
  os.system(f"cp {file_name} DiffPepBuilder/examples/receptor_data/")
else:
  file_name = "alk1.pdb"
  os.system(f"cp DiffPepBuilder/examples/receptor_data/alk1.pdb test_case/")
#@markdown - **Note**: please remove non-protein components from the PDB file and ensure that the CA atoms are present.
#@markdown - **Note**: please remove any underscores (`_`) from the receptor PDB file name, as the script interprets `_` as a delimiter.

lig_chain = None #@param {type:"string"}
#@markdown  - Chain ID of the **reference** ligand. Please set to `None` if no reference ligand is included in the PDB file.
#@markdown  The model will prioritize reference ligand information over the hotspots and motif if both are given.
hotspots = "B40, B58-59, B71-72, B87" #@param {type:"string"}
motif = "B38-42, B47-50, B52-61, B64-67, B69-77, B82-87" #@param {type:"string"}
#@markdown - **Note**: the model will prioritize hotspots over the motif when both are given. See the [paper](https://arxiv.org/abs/2405.00128) for more details.

key = os.path.splitext(file_name)[0]
data = {}
if lig_chain and lig_chain != "None":
  data['lig_chain'] = lig_chain
if hotspots and hotspots != "None":
  data['hotspots'] = hotspots.replace(",", "-")
if motif and motif != "None":
  data['motif'] = motif.replace(",", "-")

json_file_write_path = "test_case/de_novo_cases.json"
final_data = {key: data}
with open(json_file_write_path, 'w') as file:
  json.dump(final_data, file, indent=4)
```

--------------------------------

### Extract SSBLIB Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Unzip the required SSBLIB data files before running design protocols.

```bash
cd SSbuilder
tar -xvf SSBLIB.tar.gz
```

--------------------------------

### Preprocess Receptor and Peptide Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/examples/DiffPepDock_demo.ipynb

Executes the batch processing script to prepare docking cases and generate ESM embeddings.

```python
#@title ### Preprocess receptor and peptide sequence data
!python DiffPepBuilder/experiments/process_batch_dock.py \
  --pdb_dir test_case \
  --write_dir test_case \
  --receptor_info_path test_case/docking_cases.json \
  --peptide_seq_path test_case/peptide_seq.fasta
```

--------------------------------

### Preprocess Receptor Data

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Processes the receptor PDB files using the provided JSON configuration for residue identification.

```bash
python experiments/process_receptor.py --pdb_dir examples/receptor_data --write_dir data/receptor_data --receptor_info_path examples/receptor_data/de_novo_cases.json
```

--------------------------------

### Split Dataset

Source: https://github.com/yuzhewangpku/diffpepbuilder/blob/main/README.md

Divides the processed dataset into training and validation sets based on the specified number of validation samples.

```bash
python experiments/split_dataset.py --input_path data/complex_dataset/metadata.csv --output_path data/complex_dataset --num_val 200
```
Tokens:
Raw