Is it suitable for protein-protein docking?

No, DiffDock is specifically trained for small molecule ligands. For protein-protein interactions, researchers should use alternative tools like AlphaFold-Multimer or DiffDock-PP.

Does DiffDock predict binding affinity?

No, DiffDock predicts 3D binding poses and confidence scores. For affinity (ΔG) or binding constant (Kd) predictions, you should combine DiffDock outputs with specialized scoring functions like GNINA or MM/GBSA.

How do I interpret confidence scores?

Scores > 0 indicate high confidence, -1.5 to 0 are moderate, and below -1.5 are low. Note that confidence measures the model's certainty about the structural pose, not the physical binding strength.

Can I run this on a CPU?

While DiffDock can run on a CPU, a GPU is strongly recommended for practical use as it typically provides a 10-100x speedup, making virtual screening tasks feasible.

What inputs are required for docking?

You need a protein target (provided as a PDB file or amino acid sequence) and a ligand (provided as a SMILES string or a structure file like SDF or MOL2).

DiffDock Molecular Docking

Name: DiffDock Molecular Docking
Author: K-Dense-AI

byK-Dense-AI

•

3,718

•

データサイエンスとML

Predicts 3D protein-ligand binding poses and confidence scores using state-of-the-art diffusion models for structure-based drug design.

DiffDock is a specialized Claude Code skill for computational chemistry and drug discovery that leverages diffusion-based deep learning to model protein-ligand interactions. It enables researchers to predict high-accuracy 3D binding poses from either protein structures (PDB) or sequences (ESMFold) alongside chemical SMILES strings. Beyond simple pose prediction, the skill provides confidence scoring to assess reliability and supports high-throughput virtual screening of compound libraries, bridging the gap between raw chemical data and structural insights for lead optimization.

主な機能

013,718 GitHub stars

02Batch processing for large-scale virtual screening

03Automated confidence scoring for prediction reliability

04High-accuracy 3D ligand binding pose prediction

05Support for protein structures (PDB) and sequences (ESMFold)

06Integration with ligand formats including SMILES, SDF, and MOL2

ユースケース

01Virtual screening of chemical libraries against protein targets

02Structure-based drug discovery and lead optimization

03Binding site prediction and molecular interaction analysis

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add k-dense-ai/claude-scientific-skills diffdock

For use in Claude.ai and ChatGPT

Download Skill