Is a GPU required to use this skill?

While DiffDock can run on a CPU, a GPU is strongly recommended. Using a GPU typically provides a 10x to 100x speedup, which is essential for virtual screening of multiple compounds.

Can I run docking using only a protein sequence?

Yes, this skill supports protein sequences by utilizing ESMFold to generate embeddings, allowing for docking even when a high-resolution PDB structure is unavailable.

Does DiffDock predict binding affinity?

No, DiffDock is specifically designed to predict the 3D binding pose and providing a confidence score regarding the structural accuracy. For affinity prediction (ΔG), you should combine DiffDock results with scoring functions like GNINA or MM/GBSA.

What is considered a 'good' confidence score in DiffDock?

Confidence scores greater than 0 are generally considered high confidence. Scores between -1.5 and 0 are moderate, while scores below -1.5 indicate low prediction certainty.

DiffDock Molecular Docking

Name: DiffDock Molecular Docking
Author: Microck

byMicrock

•

データサイエンスとML

Predicts high-accuracy 3D protein-ligand binding poses using diffusion-based deep learning for structure-based drug design.

概要

This skill integrates DiffDock, a state-of-the-art diffusion model, into the Claude environment to automate the prediction of molecular binding poses. It enables researchers and bioinformaticians to process protein structures or sequences against compound libraries, generating ranked 3D coordinates and reliability confidence scores. Whether performing single-complex docking or large-scale virtual screening, this skill streamlines the computational pipeline for early-stage drug discovery and structural biology analysis.

主な機能

Supports both protein structures (PDB) and amino acid sequences (ESMFold)
81 GitHub stars
Handles diverse ligand inputs including SMILES strings and SDF files
Generates reliability confidence scores for predicted poses
Predicts 3D binding poses with high accuracy using diffusion models
Processes single complexes or batch virtual screening campaigns

ユースケース

Structure-based drug design and lead optimization
Virtual screening of compound libraries against specific protein targets
Predicting ligand binding sites and structural poses for chemical biology

概要

主な機能

Supports both protein structures (PDB) and amino acid sequences (ESMFold)
81 GitHub stars
Handles diverse ligand inputs including SMILES strings and SDF files
Generates reliability confidence scores for predicted poses
Predicts 3D binding poses with high accuracy using diffusion models
Processes single complexes or batch virtual screening campaigns

ユースケース

Structure-based drug design and lead optimization
Virtual screening of compound libraries against specific protein targets
Predicting ligand binding sites and structural poses for chemical biology