Skip to content
Get packs

DiffDock Molecular Docking

Guides an agent through predicting how small molecules bind to proteins using DiffDock, with batch runs and pose triage.

Source Download skill
Repository
K-Dense-AI/scientific-agent-skills
Stars
55,780
Licence
MIT
Files
9 files

What it does

Once installed, the agent knows how to set up the upstream DiffDock docking tool, prepare protein and ligand inputs, run single or batch predictions, and read the ranked poses and confidence scores. It also warns about common pitfalls, such as treating confidence as binding affinity or reusing output folders. The skill itself does not run docking; it relies on a separate DiffDock installation (or Docker image) that must be set up first.

Great for

  • Computational chemists or structural biologists who want to generate candidate ligand poses for a protein target
  • Researchers preparing reproducible batch docking or ensemble runs across receptor structures
  • Students learning how molecular docking workflows and confidence scores work

Skip it if

  • Predicting binding affinity or ranking drugs by potency: the skill explicitly says DiffDock confidence is not affinity and should not be used for hit identification on its own.
  • Clinical or safety decisions, or anything where docking output is treated as proof of biological activity, because predictions are unvalidated computational guesses.
  • Users without a CUDA-capable GPU or the ability to install a large, pinned Python environment, since the upstream setup is heavy and sequence folding requires CUDA.

Try asking

  • “Dock aspirin's SMILES into my protein.pdb receptor and show me the top 10 ranked poses.”
  • “Build a batch CSV of five ligands against one receptor and check it for errors before running.”
  • “Summarise the confidence scores from my DiffDock results folder into a CSV.”
  • “Set up the DiffDock v1.1.3 environment and check that my install has the required packages.”
Step-by-step guide: how to use DiffDock Molecular DockingInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 9 original files, the MIT licence and an attribution note.

    diffdock.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip diffdock.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip diffdock.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Use a fresh output directory for every run to avoid mixing old rank files with new results.
  • Put any sampling settings on the command line only after checking the YAML config, because YAML values overwrite matching CLI flags.
  • Read the confidence and limitations reference before interpreting any ranking, and validate top poses with independent methods.

What's inside

  • assets/batch_template.csv
  • assets/custom_inference_config.yaml
  • references/confidence_and_limitations.md
  • references/parameters_reference.md
  • references/workflows_examples.md
  • scripts/analyze_results.py
  • scripts/prepare_batch_csv.py
  • scripts/setup_check.py
  • SKILL.md

Redistributed unmodified from K-Dense-AI/scientific-agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.