Use case · Idea to test
Molecule and drug research
Rank research candidates against a stated goal.
- Candidate structures
- Score desired properties
- Choose what to test
A research tool could help shortlist molecular candidates for further evaluation. The older site proposed drug discovery; this remains a research application, not a demonstrated treatment-design system.
Where the small model fits
A domain model represents each candidate. A small scorer learns from measured properties or expert comparisons, while separate checks enforce the experiment's constraints.
What would need to work
Compare selection against conventional search using independently measured outcomes and the same evaluation budget. A high learned score does not establish safety, efficacy or biological activity. No molecular benchmark has been run here.
Try the related work
This demo learns from simple drawings. It does not run the application described above.