
Description
diffdock is an agent skill for predicting protein and small-molecule binding poses with DiffDock and DiffDock-L. You can start from a PDB structure or a protein sequence, and a ligand given as SMILES, SDF or MOL2. The model generates candidate ligand poses and ranks them by model confidence. The skill is clear that confidence is neither a measured probability of correctness nor a binding affinity. It covers a single protein-ligand pair, batch docking, and runs against separate receptor conformations. It advises treating a library-wide sort by confidence as pose triage, not as hit identification. The skill targets DiffDock v1.1.3. Its source and the current main branch's inference script were checked on 30 September 2026. Bundled helpers were tested on tiny synthetic inputs, but pretrained docking, ESMFold, CUDA, Docker, GNINA and hosted demo runs were not executed in that review. The skill labels the commands that need those components as source-verified recipes rather than demonstrated end-to-end runs. It also includes steps for setting up the upstream environment. It is part of the k-dense-ai scientific agent skills repository on GitHub, which has about 49,600 stars.…
Highlights
- Ligand poses from PDB or sequence plus SMILES, SDF or MOL2
- DiffDock and DiffDock-L
- Batch docking and pose triage
- Confidence is not a probability or affinity
- Targets DiffDock v1.1.3
Ratings & Reviews
No ratings yet
Be the first to rate diffdock.
People also visit
diffdock FAQ
What does DiffDock predict?
Binding poses for a small molecule in a protein, which it ranks by model confidence.
Does a high confidence mean strong binding?
No. The skill says confidence is neither a measured probability of correctness nor binding affinity.
Can it screen a library?
It can run batches, but the skill says a library-wide sort by confidence is pose triage, not hit identification.
Which inputs does it accept?
A PDB structure or a sequence for the protein, and SMILES, SDF or MOL2 for the ligand.
Which version does it target?
DiffDock v1.1.3. Some components such as CUDA, Docker and GNINA were not run in the skill's review.
How do I install it?
Run npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill diffdock.
What are the best diffdock alternatives?
Popular alternatives to diffdock include esm, geniml and pytdc. See all diffdock alternatives
