Use case · Idea to test
Protein research
Score candidates without treating scores as proof.
- Candidate representations
- Rank a research objective
- Check measured outcomes
The earlier whitepaper proposed protein design as another domain for candidate scoring. Here it remains an untested research idea with no biological results claimed.
Where the small model fits
A suitable domain model would provide representations. A small head could rank a defined objective, provided relevant training labels and an independent evaluation were available.
What would need to work
Evaluate against measured outcomes on unseen candidates. Structural plausibility or a learned score is not evidence that a proposed protein performs a desired function.
Try the related work
This demo learns from simple drawings. It does not run the application described above.