Small-molecule screening
Generates candidates from a seed compound and docks them one at a time against a receptor until enough actually beat a real affinity bar. Use once a receptor exists and the question is "which of these would actually work," not just "which of these look drug-like."
Small-molecule screening
What it does
Connects two tools this platform already has, closing a real gap between them. Small-molecule generation proposes candidates from a seed but has no way to say which ones are any good; Small-molecule docking scores exactly one candidate at a time. This tool generates a full candidate set from your seed, then docks candidates one at a time — in the same order Small- molecule generation would rank them — until enough clear a real affinity bar, or the candidate pool runs out.
Only real hits become saved evidence. Every candidate tried is recorded either way, but a candidate that doesn't beat the bar never becomes a saved compound cluttering your project — only your seed's own docked pose (a baseline for comparison) and each candidate that actually beat it do.
How screening decides what counts
Your seed is always docked first, against the same receptor, to establish a real baseline affinity — not a heuristic guess. By default, a candidate "hits" if it docks better (a more negative affinity) than that baseline; you can set your own affinity threshold instead if you have one in mind. Candidates are tried in the order Small-molecule generation would already rank them (best drug-likeness first) — a reasonable starting order, not a prediction of which will actually dock well; only the real docking result decides that.
Screening stops the moment enough candidates have hit, or once every generated candidate has been tried. Finding fewer hits than you asked for is still a real, complete result — not an error — if the candidate pool ran out first.
Reading the output
The Result panel and the run's own record list every candidate tried, hit or not, with its docking affinity next to your seed's baseline — the full picture of what was actually tried, separate from what became evidence. Only the seed's baseline pose and each hit's own docked pose are saved as compounds you can use elsewhere on the platform.
A hit's docked pose is an ordinary docking result: feed it into Molecular dynamics (as the ligand, alongside the same receptor) to check whether the binding pose actually holds up, or into ADMET prediction to check its absorption/distribution/metabolism/excretion/toxicity profile — no extra steps needed, it's the same file shape either tool already accepts.
Input
One seed compound, saved as SDF, and one receptor, saved as PDB or mmCIF.
How long this takes
Real measurement, not a guess. Docking one candidate takes about a minute at this tool's default search thoroughness (exhaustiveness); across a real generated candidate set, roughly one in five candidates beat their seed's own baseline. Put together, finding the default target of 5 hits typically takes very roughly 12 to 60 minutes — but this genuinely varies by seed and receptor, and there's no way to know in advance which you'll get. Worst case, if the full candidate pool never clears the bar: around 8 hours. Runs on this platform's own compute queue; a run's progress (candidates tried, hits found so far) shows in Run Results.
Important limitations
Blind docking, same as Small-molecule docking. The search covers the whole receptor, not a chosen pocket — see that tool's own documentation for what this trades off.
Shares its compute queue with every other Small-molecule docking and screening run on this deployment, one job at a time. A long screening run can leave someone else's quick single-candidate docking run waiting behind it.
QED and SA score, shown alongside each candidate, are heuristics, not predictions — see Small-molecule generation's own documentation for the same caveat. They decide the order candidates are tried in, never whether one counts as a hit.
No candidate-count or search-thoroughness auto-tuning. You set search thoroughness once; screening doesn't lower it automatically as the candidate pool runs low or raise it because hits are coming easily.
Starts from one seed, not an uploaded list of candidates. This tool generates its own candidate set from a seed compound — it isn't a way to screen an arbitrary list you already have.