science1 publisher
Contrastive training lands a protein's sequence and its structure in the same spot on the map
Reported in PNAS, the CLSS model reproduces relationships from the expert-curated ECOD and CATH classifications it never saw in training, and it can place short protein fragments beside complete sequences and structures.
Publishers:phys.org
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+20
- Incentives58
- Confidence55