Google DeepMind shows Antigravity as a science workbench
An agent finds and characterizes a candidate bacteriocin against Streptococcus pseudopneumoniae (wet-lab tests pending)
Confirmed
The takeaway
A DeepMind blog post walks through an Antigravity session (Gemini 3.8 Flash) that searched UniProt, annotated proteins with ProtNLM, checked novelty with BLAST, compared AlphaFold structures and genomic context, and proposed a mechanism and lab plan for a candidate antimicrobial peptide. No experimental results yet.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 15 October 2026
Your AI and this story
- GPT-6 Astra168 days after its cutoff
- Claude Opus 5.5107 days after its cutoff
- Gemini 4 Argon198 days after its cutoff
- Grok 4.7137 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 107 days before it.
Key facts
- Authors: Pengfei Tian, Maxwell L. Bileschi, Zoi Tsangalidou, Sukhdeep Singh, Lucy J. Colwell (Google DeepMind); model selected in screenshots: Gemini 3.8 Flash
- Candidate: UniProt A0A0U0D591, 97 amino acids, Lactococcin 972 family, from Streptococcus pseudopneumoniae (respiratory infections)
- Pipeline inside one agent session: UniProt filtering → ProtNLM function prediction → BLAST against Swiss-Prot (EMBL-EBI) → AlphaFold model, remote structural homologs and superposition → genomic context (ABC transporter permease, ATP-binding partner, second bacteriocin, XRE regulator) → literature review → plan for expression, susceptibility, target-engagement and toxicity tests
- Status: ‘the lead and several other candidates’ from the pipeline are in wet-lab validation; no hit rates or experimental data; no paper or code linked
- New tool shown: a Biomolecular Structure Viewer embedded in Antigravity, a shared 3D workspace where researcher and agent annotate structures
What happened
DeepMind published a step-by-step account of one agentic research session in Antigravity, Google’s agent IDE, used here for biology rather than code. The agent searched protein databases, called domain tools (ProtNLM, BLAST, the AlphaFold database), compared structures in an embedded 3D viewer and wrote up a mechanism hypothesis and an experimental plan for a bacteriocin-like peptide.
Why it matters
It shows Google positioning Antigravity as a general science workbench. The output is a hypothesis, though: no wet-lab result is reported yet, so the post demonstrates a workflow, not a discovery.
Sources
1 source from 1 site. Numbers match the chips in the text.
1 source: 1 primary
Primary
- Google DeepMind blog: Accelerating the search for antimicrobial peptides with Google Antigravitydeepmind.google, official
Changes
- Filed (deepmind.google)