Nature·· 12 hours agoPickScore86
Programmable design of functional proteins from natural language.
In brief
A Nature paper introduces Pinal, a 16-billion-parameter foundation model that generates candidate protein sequences from natural-language descriptions of desired function. Trained on 1.7 billion synthetically annotated protein-text pairs, the model produced candidates that, in computational evaluations, combined high predicted foldability, functional-description alignment and sequence diversity.
Why it matters
Uses natural language to specify protein function and validates generated results in four design tasks, providing a reference for testing whether this design paradigm transfers to new functions.
Source: Nature · doi.org