skills.rest
silverstein
silversteinCOMMUNITY
Explore all skills by this author →
Website𝕏@matsilversteinGitHub

diffdock

Predict ranked protein–ligand binding poses with confidence scores from PDB structures or sequences and SMILES inputs.

21|2|Updated Dec 8, 2025
Search Tags:#virtual screening#molecular docking#diffusion models#confidence scoring#protein-ligand#structure-based drug design
One-click install
npx skills add https://github.com/silverstein/claude-scientific-skills-desktop --skill diffdock-silverstein
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: diffdock
Source: https://github.com/silverstein/claude-scientific-skills-desktop/tree/main/corpus/diffdock
Command: npx skills add https://github.com/silverstein/claude-scientific-skills-desktop --skill diffdock-silverstein

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pandas, rdkit, os, re, argparse, sys, json, pathlib, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill helps you predict plausible 3D binding poses of small molecules in protein binding sites, including an internal confidence score to guide which poses to trust. It addresses the bottleneck of running modern structure-based docking workflows without manually stitching together multiple steps and parameters.

Core Features & Use Cases

  • Diffusion-based pose prediction: Generates ranked ligand binding poses from a protein structure (PDB) and ligand input (SMILES or common structure formats).
  • Protein input flexibility: Accepts either PDB files or protein sequences (via ESMFold folding) as protein targets.
  • Confidence scoring for ranking: Produces confidence scores that reflect prediction certainty (not binding affinity), helping prioritize downstream rescoring/validation.
  • Virtual screening and batching: Supports single complexes or batch docking using a CSV input workflow for screening compound libraries.
  • Result analysis and troubleshooting: Provides guidance for interpreting confidence and common failure modes, plus scripts for parsing outputs.

Quick Start

Run DiffDock on one protein and ligand by asking the AI to execute the single-complex inference workflow with a protein PDB file and a ligand SMILES string, saving outputs to a results directory.

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 from a PDB structure and SMILES string?▼

Protein-ligand binding poses are predicted using diffusion models that process a PDB structure and ligand SMILES to generate multiple ranked 3D poses. You can execute the single-complex inference workflow to output these predictions directly to a results directory.

Can I use a protein sequence instead of a PDB file for molecular docking?▼

Molecular docking supports protein sequences as input by leveraging ESMFold folding to generate the required structure. This allows you to run diffusion-based pose prediction even when an existing PDB file is unavailable for your target.

Does the confidence score predict ligand binding affinity?▼

The confidence score reflects prediction certainty for the generated poses, not binding affinity prediction. It helps prioritize which docking poses to trust for downstream rescoring and validation tasks.

How do I run virtual screening for a compound library?▼

Virtual screening is executed through a batch docking workflow driven by a CSV batch file. This process supports ensemble docking across multiple protein conformations and generates ranked poses with confidence scores for the entire library.

What is the best way to interpret multiple ranked docking poses?▼

Interpreting multiple ranked poses involves using the provided confidence scores to gauge prediction certainty. The Skill includes scripts for parsing outputs and guidance on common failure modes to support confidence-aware interpretation.

Why does my batch virtual screening workflow require a specific input format?▼

Batch virtual screening requires a CSV input format to systematically map ligand SMILES and protein targets for the diffusion models. This structure ensures all compounds are processed correctly and ranked poses are generated consistently.

Related Skills

Explore More Education & Research Skills→
qinyan-aiqinyan-ai
48

diffdock

Predict ranked 3D binding poses for protein-ligand complexes from PDB and SMILES inputs.

Community
Advanced
OwnLabAIOwnLabAI
21

diffdock

Generate protein-ligand docking poses with diffusion models and confidence scores.

Official
Advanced
biomaps-infrabiomaps-infra

bio-diffdock

Predict molecular docking poses for protein-ligand interactions using diffusion models.

Official
Advanced
LeonChaoXLeonChaoX
783

diffdock

Predict ranked 3D protein–ligand binding poses with confidence scores.

Community
Advanced