Small-molecule generation
Proposes new candidate molecules by extending a seed compound at an open valence, using CReM over a real ChEMBL fragment database. Use to propose a scaffold nobody has drawn yet, when Similarity search's existing library isn't enough.
Small-molecule generation
What it does
Takes one seed compound and proposes new candidates by adding a fragment at an open position — real chemistry, not a guess: every fragment it adds was observed being attached in exactly that kind of local chemical environment somewhere in ChEMBL.
Each candidate comes back with two numbers to help triage: QED (drug-likeness, 0 to 1, higher is more drug-like) and SA score (synthetic accessibility, 1 to 10, lower is easier to make). Results are ranked by QED, best first.
How candidates are generated
The underlying method (CReM) works by fragment-context matching: it looks at the chemical neighborhood right around an open position on the seed, and finds fragments from a large reference database (built from ChEMBL) that were seen attached in an identical neighborhood elsewhere. Because every candidate is built this way, an invalid molecule is essentially impossible — confirmed at very large scale during evaluation, not just expected.
Reading the output
The ranked list, plus each candidate's QED and SA score, are what to read. A high QED candidate with a low (easy) SA score is the strongest starting point for follow-up work; a high QED candidate with a high SA score is worth a second look before committing time to it.
Input
One seed compound, saved as an SDF.
Important limitations
This tool proposes candidates from a seed you give it — it doesn't take a binding pocket. If the goal is "design something for this specific pocket," the intended path is: generate candidates here, then dock and score them with the Small-molecule docking tool — not a single tool that does both.
QED and SA score are heuristics, not predictions. QED estimates how "drug-like" a molecule's properties look based on known approved drugs; it is not a prediction of biological activity. SA score estimates how hard a molecule would be to synthesize based on the complexity of fragments it's built from; it is not a real synthesis route.