GUIDE · MODELS · #2057
Google Antigravity and ProtNLM identify novel antimicrobial peptide candidate in UniProt
Google describes using its agentic science workbench Antigravity together with ProtNLM, a protein-language model, to annotate millions of unreviewed UniProt sequences and — in a technical walkthrough — prioritize a 97‑amino‑acid bacteriocin (UniProt ID A0A0U0D591, Lactococcin 972 family) predicted to have activity against respiratory pathogens; the lead is undergoing wet‑lab validation. The Antigravity workflow combined sequence-function predictions, BLAST/Swiss‑Prot checks, and AlphaFold structural comparisons in a single interactive session.
KEY POINTS
- Google describes using its agentic science workbench Antigravity together with ProtNLM, a protein-language model, to annotate millions of unreviewed UniProt sequences and — in a technical walkthrough — prioritize a 97‑amino‑acid bacteriocin (UniProt ID A0A0U0D591, Lactococcin 972 family) predicted to have activity against respiratory pathogens; the lead is undergoing wet‑lab validation.
- The Antigravity workflow combined sequence-function predictions, BLAST/Swiss‑Prot checks, and AlphaFold structural comparisons in a single interactive session.
- This shows a production-scale protein language model and an agentic workbench can compress multi-database discovery workflows and surface novel antimicrobial candidates from uncharacterized UniProt sequences, potentially accelerating antibiotic discovery pipelines.
WHY IT MATTERS
This shows a production-scale protein language model and an agentic workbench can compress multi-database discovery workflows and surface novel antimicrobial candidates from uncharacterized UniProt sequences, potentially accelerating antibiotic discovery pipelines.