diffdock

Name: diffdock
Availability: InStock
Author: zeyuyang-0420

Predict 3D binding poses of small molecule ligands to protein targets.

Updated May 8, 2026

Search Tags:#deep learning #protein structure #drug discovery #molecular docking #ligand binding

Ready to run

npx skills add https://github.com/Zeyuyang-0420/bio-ai-research-skills --skill diffdock-zeyuyang-0420

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: diffdock
Source: https://github.com/Zeyuyang-0420/bio-ai-research-skills/tree/main/categories/drug-discovery-molecular-modeling/diffdock
Command: npx skills add https://github.com/Zeyuyang-0420/bio-ai-research-skills --skill diffdock-zeyuyang-0420

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, torch-geometric, rdkit, esmfold, diffdock, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill solves the problem of predicting the 3D binding poses of small molecule ligands to protein targets, which is crucial for structure-based drug discovery and chemical biology.

Core Features & Use Cases

High Accuracy Pose Prediction: Using deep learning, it predicts ligand binding poses with high accuracy.
Supports Protein Structures and Sequences: Process protein structures (PDB files) or sequences (via ESMFold).
Batch Virtual Screening: Process multiple protein-ligand pairs in batch mode for virtual screening campaigns.
Confidence Scores: Provides confidence scores to assess prediction reliability.
Use Cases: Ideal for "Dock this ligand to a protein" or "predict binding site" tasks, and for structure-based drug design or lead optimization.

Quick Start

Use the diffdock skill to dock a ligand to a protein target with the following command:

python -m inference --protein_path protein.pdb --ligand "CC(=O)Oc1ccccc1C(=O)O" --out_dir results/single_docking/

Frequently Asked Questions about diffdock

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I predict protein-ligand binding poses for structure-based drug discovery?▼

To predict protein-ligand binding poses, you can use deep learning models to dock small molecule ligands to protein targets. This approach provides high accuracy 3D binding pose predictions and confidence scores for structure-based drug discovery.

Can I run molecular docking if I only have a protein sequence instead of a PDB file?▼

Yes, you can run molecular docking with only a protein sequence. The tool supports processing protein sequences via ESMFold to generate the required structures, alongside accepting standard PDB files for ligand binding pose prediction.

How do I perform batch virtual screening for multiple protein-ligand pairs?▼

You can perform batch virtual screening by processing multiple protein-ligand pairs in batch mode. This allows you to evaluate numerous small molecule ligands against protein targets efficiently for lead optimization campaigns.

Does molecular docking with deep learning provide confidence scores for binding predictions?▼

Molecular docking with deep learning provides confidence scores to assess prediction reliability. These scores help evaluate the accuracy of predicted 3D binding poses for small molecule ligands during virtual screening.

What dependencies do I need to run deep learning molecular docking?▼

To run deep learning molecular docking, you need dependencies including torch, torch-geometric, rdkit, esmfold, and diffdock. These tools facilitate protein structure processing and ligand binding pose prediction.

What is the best way to dock a ligand to a protein target using a command line interface?▼

The best way to dock a ligand to a protein target is by running a python inference command with specified protein path and ligand parameters. This outputs predicted 3D binding poses and confidence scores to a designated directory.