OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Pepglad

PepGLAD is Python research code implementing geometric latent diffusion for full-atom peptide sequence-structure co-design and peptide binding-conformation generation. The repository includes trained checkpoints and instructions for inference and reproducing paper experiments.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Pepglad — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use PepGLAD when a research workflow needs local model code for peptide sequence-structure co-design or binding-conformation generation, and the team can prepare protein structures and a pocket description.

Who it is for

  • Researchers exploring generative peptide design against protein binding sites.
  • Developers building research workflows that use peptide sequence and structure generation.
  • Teams reproducing or evaluating the repository's peptide-design experiments.

Documented uses

  • Generate peptide sequence-and-structure candidates for a target protein pocket.
  • Generate binding conformations for a supplied peptide sequence and target pocket.
  • Set up the provided checkpoints for inference without downloading benchmark datasets.
  • Reproduce the paper's model-training, generation, and evaluation workflow.
Before you choose

Requirements & limits

The code is publicly available on GitHub under the MIT license. Trained weights are linked from the repository's v1.0 release. The documented inference workflow requires downloading weights; the benchmark datasets are optional for inference.

  • The repository offers separate modes for sequence-structure co-design and fixed-sequence conformation generation.
  • Inference needs a protein PDB structure and a pocket description; the README allows pockets to be detected from chains or prepared manually.
  • The README reports CUDA 11.7 and PyTorch 1.13.1 as tested versions.
  • PyRosetta is optional and is used for calculating interface energy.
  • Benchmarking datasets are optional when using trained weights for inference.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Generate peptide sequence-structure candidates

Use the documented codesign workflow to generate peptide candidates for a target binding pocket.

Before you start

  • PepGLAD's Conda environment and downloaded trained weights.
  • A target protein structure in PDB format.
  • A binding pocket detected from target and ligand chains or prepared manually.

Steps

  1. Create a pocket description with the command below, providing the PDB file, target chains, ligand chains, and output path as shown in the README.

    Documented command
    python -m api.detect_pocket

  2. Run the command below in codesign mode with the PDB and pocket files, an output directory, length bounds, and a requested sample count.

    Documented command
    python -m api.run

Expected outcome

The documented example requests ten peptide sequence-structure generations with lengths from 8 up to, but not including, 15 residues, saved in the chosen output directory.

Guide limitations

  • The README's example uses CUDA and an example PDB; it does not establish compatibility across hardware or inputs.
  • Generated candidates are model outputs and do not establish biological activity or binding.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
THUNLP-MT/PepGLAD · Official homepage title · retained in its source language
How we review sourcesSuggest a correction