What Liganx is, how far to trust it, and what happens to your data — the things worth knowing before you run your first job. If your question isn't here, get in touch.
- Is Liganx free?
- Yes — Liganx is free for academic use, and you can run a docking job without creating an account at all. Signing in with an email keeps your job history and raises your free-run limit. Free jobs run on GPU-accelerated QuickVina2, with every pose checked by PoseBusters, a Vinardo re-score, and RDKit strain analysis. GNINA CNN rescoring and Boltz-2 co-folding run alongside on the same job for Pro users.
- What does “mutation-aware” docking actually mean?
- Most docking tools dock a compound against one protein structure. Liganx docks it against the wild-type protein and one or more clinically relevant mutants in parallel, then ranks the change in predicted binding — the Δ-score — between them. That tells you where a compound is likely to hold and where a resistance mutation is likely to weaken it, which is the question that actually matters in the resistance setting. Read the mutation-docking guide →
- How accurate is it? Can I trust the scores?
- We publish the full benchmark instead of asking you to take a slogan on faith. On the Astex Diverse Set, GNINA reaches 69% top-1 pose accuracy under 2 Å, fully automated. Boltz-2 poses carry a model-confidence badge (the model's own ligand pLDDT, binned green / amber / red): on a 30-target holdout, green poses were right 12 of 13 times, and low-confidence poses are flagged. We also re-run eleven literature-anchored resistance cases with public verdicts, including the ones our method gets wrong and the structural reason why. See the full validation report →
- What happens to my compounds? Is my data private?
- Liganx is built for proprietary work. You can run a job without an account, we run no third-party analytics trackers and set no cookies beyond your dark/light theme choice, and no data leaves North America. Your receptor and ligand files are sent to a GPU worker only for the duration of a single docking call and discarded there afterward. Job records are keyed to a random share link — treat that link like a password for confidential work — and we will delete any job from our database on request. Read the full privacy policy →
- How should I read the Δ-scores? What is the noise floor?
- Vina/QuickVina2 scoring carries roughly ±1 kcal/mol of noise at default exhaustiveness. So a Δ larger than about 1 kcal/mol is interpretable, while smaller shifts live near the noise floor and shouldn't be over-read. Liganx puts a “within-noise” badge on any Δ inside that band, so you don't chase a 0.3 kcal/mol difference as if it were a real signal.
- Do I need to install anything?
- No. Liganx runs entirely in the browser — no PyMOL, no FoldX setup, no AutoDock wrangling. Vina, GNINA CNN rescoring, and Boltz-2 co-folding all run on GPU on our infrastructure and stream the results back to the page.
- Which targets and mutations are supported?
- You can pick from a curated library of clinically actionable kinases with pre-defined pocket boxes, or upload your own PDB. Mutations come from a curated library — click EGFR T790M, KRAS G12C, or BRAF V600E — or type your own. Mutations that fall outside the binding pocket are flagged rather than given a possibly-misleading score.
- How long does a run take, and how many compounds can I screen?
- A standard mutation-aware run returns in minutes. You can screen up to 1,000 compounds against a (target, mutation) pair in a single pass and get a hit list ranked by selectivity, then promote the top hits to a full job in one click. For an instant look, our pre-computed screenings of FDA-launched oncology inhibitors load in about a second. Browse pre-computed screenings →
- How is Liganx different from Schrödinger or plain AutoDock Vina?
- It sits in the missing middle. Free servers give you raw docking with no validation and no mutation workflow; Schrödinger is powerful but expensive and desktop-bound. Liganx is browser-based, mutation-aware by default, ships a published and reproducible validation report, and packages the whole resistance question — wild-type versus a full variant panel — in a way no other self-serve tool does.
- Can I use Liganx for commercial or industry work?
- The hosted app is a research preview, and one component — FoldX, used for higher-fidelity mutant building — is free for academic use but requires a commercial license for industry workflows under its own terms. If you want to use Liganx in a commercial setting, get in touch and we'll walk through what applies to your case. Contact us →
- What are the method's known limitations?
- We publish them on purpose. Docking is a fast first pass, not a final verdict. Liganx docks against one rigid receptor conformation per structure, so it can miss effects that require the protein to change shape — the EGFR L858R activating mutation is a documented case it gets wrong. The hosted mutant is built with PDBFixer's residue substitution, which sets the new identity but doesn't energy-minimize, so drastic side-chain changes can add clash signal to the Δ. Mutations outside the pocket are flagged, not scored. See our methods and limitations →