You can deploy this model and start earning money today!

Irwiny/
PepGLAD

Views

No views yet

PepGLAD: Full-Atom Peptide Design with Geometric Latent Diffusion

Quick Links

Setup
Usage
- Peptide Sequence-Structure Co-Design
- Peptide Binding Conformation Generation
Reproduction of Paper Experiments
- Codesign
- Binding Conformation Generation
Contact
Reference

Updates

Changes for compatibilities and extended functionalities are saved in beta branch. Thank @Barry0121 for the help.

pyTorch 2.6.0 and openmm 8.2.0 are supported, with new environment configure at 2025_env.yaml.
Support non-canonical amino acids in detect_pocket.py.

Setup

Environment

The conda environment can be constructed with the configuration env.yaml:

conda env create -f env.yaml

The codes are tested with cuda version 11.7 and pytorch version 1.13.1.

Don't forget to activate the environment before running the codes:

conda activate PepGLAD

(Optional) pyRosetta

PyRosetta is used to calculate interface energy of generated peptides. If you are interested in it, please follow the instruction here to install.

(Optional) Datasets

These datasets are only used for benchmarking models. If you just want to use the trained weights for inferencing on your cases, there is no need to download these datasets.

PepBench

Download

The datasets, which are originally introduced in this paper, are uploaded to Zenodo at this url. You can download them as follows:

bash

mkdir datasets  # all datasets will be put into this directory
wget https://zenodo.org/records/13373108/files/train_valid.tar.gz?download=1 -O ./datasets/train_valid.tar.gz   # training/validation
wget https://zenodo.org/records/13373108/files/LNR.tar.gz?download=1 -O ./datasets/LNR.tar.gz   # test set
wget https://zenodo.org/records/13373108/files/ProtFrag.tar.gz?download=1 -O ./datasets/ProtFrag.tar.gz     # augmentation dataset

Decompresss

bash

tar zxvf ./datasets/train_valid.tar.gz -C ./datasets
tar zxvf ./datasets/LNR.tar.gz -C ./datasets
tar zxvf ./datasets/ProtFrag.tar.gz -C ./datasets

Process

bash

python -m scripts.data_process.process --index ./datasets/train_valid/all.txt  --out_dir ./datasets/train_valid/processed  # train/validation set
python -m scripts.data_process.process --index ./datasets/LNR/test.txt  --out_dir ./datasets/LNR/processed  # test set
python -m scripts.data_process.process --index ./datasets/ProtFrag/all.txt --out_dir ./datasets/ProtFrag/processed # augmentation dataset

The index of processed data for train/validation splits need to be generated as follows, which will result in datasets/train_valid/processed/train_index.txt and datasets/train_valid/processed/valid_index.txt:

python -m scripts.data_process.split --train_index datasets/train_valid/train.txt --valid_index datasets/train_valid/valid.txt --processed_dir datasets/train_valid/processed/

PepBDB

Download

wget http://huanglab.phys.hust.edu.cn/pepbdb/db/download/pepbdb-20200318.tgz -O ./datasets/pepbdb.tgz

Decompress

tar zxvf ./datasets/pepbdb.tgz -C ./datasets/pepbdb

Process

bash

python -m scripts.data_process.pepbdb --index ./datasets/pepbdb/peptidelist.txt --out_dir ./datasets/pepbdb/processed
python -m scripts.data_process.split --train_index ./datasets/pepbdb/train.txt --valid_index ./datasets/pepbdb/valid.txt --test_index ./datasets/pepbdb/test.txt --processed_dir datasets/pepbdb/processed/
mv ./datasets/pepbdb/processed/pdbs ./dataset/pepbdb  # re-locate

Trained Weights

codesign: ./checkpoint/codesign.ckpt
conformation generation: ./checkpoints/fixseq.ckpt

Both can be downloaded at the release page. These checkpoints were trained on PepBench.

Usage

:warning: Before using the following codes, please first download the trained weights mentioned above.

Peptide Sequence-Structure Co-Design

Take ./assets/1ssc_A_B.pdb as an example, where chain A is the target protein:

bash

# obtain the binding site, which might also be manually crafted or from other ligands (e.g. small molecule, antibodies)
python -m api.detect_pocket --pdb assets/1ssc_A_B.pdb --target_chains A --ligand_chains B --out assets/1ssc_A_pocket.json
# sequence-structure codesign with length in [8, 15)
CUDA_VISIBLE_DEVICES=0 python -m api.run \
    --mode codesign \
    --pdb assets/1ssc_A_B.pdb \
    --pocket assets/1ssc_A_pocket.json \
    --out_dir ./output/codesign \
    --length_min 8 \
    --length_max 15 \
    --n_samples 10

Then 10 generations will be outputed under the folder ./output/codesign.

Peptide Binding Conformation Generation

Take ./assets/1ssc_A_B.pdb as an example, where chain A is the target protein:

bash

# obtain the binding site, which might also be manually crafted or from other ligands (e.g. small molecule, antibodies)
python -m api.detect_pocket --pdb assets/1ssc_A_B.pdb --target_chains A --ligand_chains B --out assets/1ssc_A_pocket.json
# generate binding conformation
CUDA_VISIBLE_DEVICES=0 python -m api.run \
    --mode struct_pred \
    --pdb assets/1ssc_A_B.pdb \
    --pocket assets/1ssc_A_pocket.json \
    --out_dir ./output/struct_pred \
    --peptide_seq PYVPVHFDASV \
    --n_samples 10

Then 10 conformations will be outputed under the folder ./output/struct_pred.

Reproduction of Paper Experiments

Each task requires the following steps, which we have integrated into the script ./scripts/run_exp_pipe.sh:

Train autoencoder
Train latent diffusion model
Calculate distribution of latent distances between consecutive residues
Generation & Evaluation

On the other hand, if you want to evaluate existing checkpoints, please follow the instructions below (e.g. conformation generation):

bash

# generate results on the test set and save to ./results/fixseq
python generate.py --config configs/pepbench/test_fixseq.yaml --ckpt checkpoints/fixseq.ckpt --gpu 0 --save_dir ./results/fixseq
# calculate metrics
python cal_metrics.py --results ./results/fixseq/results.jsonl

Codesign

Codesign experiments on PepBench:

GPU=0 bash scripts/run_exp_pipe.sh pepbench_codesign configs/pepbench/autoencoder/train_codesign.yaml configs/pepbench/ldm/train_codesign.yaml configs/pepbench/ldm/setup_latent_guidance.yaml configs/pepbench/test_codesign.yaml

Binding Conformation Generation

Conformation generation experiments on PepBench:

GPU=0 bash scripts/run_exp_pipe.sh pepbench_fixseq configs/pepbench/autoencoder/train_fixseq.yaml configs/pepbench/ldm/train_fixseq.yaml configs/pepbench/ldm/setup_latent_guidance.yaml configs/pepbench/test_fixseq.yaml

Contact

Thank you for your interest in our work!

Please feel free to ask about any questions about the algorithms, codes, as well as problems encountered in running them so that we can make it clearer and better. You can either create an issue in the github repo or contact us at jackie_kxz@outlook.com.

Reference

bibtex

@article{kong2025full,
  title={Full-atom peptide design with geometric latent diffusion},
  author={Kong, Xiangzhe and Jia, Yinjun and Huang, Wenbing and Liu, Yang},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={74808--74839},
  year={2025}
}

Irwiny/PepGLAD

PepGLAD: Full-Atom Peptide Design with Geometric Latent Diffusion

Quick Links

Updates

Setup

Environment

(Optional) pyRosetta

(Optional) Datasets

PepBench

PepBDB

Trained Weights

Usage

Peptide Sequence-Structure Co-Design

Peptide Binding Conformation Generation

Reproduction of Paper Experiments

Codesign

Binding Conformation Generation

Contact

Reference

Irwiny/
PepGLAD