Google Research builds AI agents that design and run wet-lab experiments
In short: Google Research describes new work on grounding AI agents in physical lab reality, using a digital-twin framework called VERID to let agents safely plan and execute liquid-handling robot protocols. Combined with closed-loop molecular design cycles for aptamers and proteins, the agents iteratively learned from wet-lab results, improving hit rates over multiple rounds and discovering single-digit nanomolar protein binders for a cancer-related target.
This summary was generated automatically by AI from Google Research's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1VERID: a live 3D digital twin of the lab workcell tracking robot arms, well volumes, reagents and tips, built on open-source PyLabRobot
- 2VERID blocks physically or logistically invalid robot commands before execution and runs fast virtual dry-run simulations
- 3Aptamer discovery loop: agents ran 3 design-test-iteration cycles (7-10 days each), raising confirmed hit rates to 33.5% (myoglobin) and 42.1% (alpha-thrombin)
- 4Agents autonomously diagnosed wet-lab failure modes (e.g., non-specific G-quadruplex binders) and revised design rules between rounds
- 5Protein design: surgical epitope excision sped up structure-prediction folding 8.6x, nearly doubling in-silico hit rate in human-agent collaboration, and yielded single-digit nanomolar binders for FAP, a cancer-related receptor
- 6Safety measures: biosecurity screening of designed sequences, simulation-before-execution via digital twin, and required human authorization of protocols
Why it matters
This shows a concrete path for LLM-style agents to move beyond text and code generation into physically grounded, closed-loop experimentation with real safety guardrails - relevant to anyone building agentic systems that must interact with the physical world or validate hypotheses against real-world feedback, not just a biology niche.
Sources
- Google ResearchOfficialPrimary sourceOriginal article →„Grounding AI agents in biological reality: Building blocks for autonomous wet labs“16 Oct 2026, 21:11
- Published by source
- 16 Oct 2026, 21:11
- Found by our system
- 16 Oct 2026, 21:12
- Summary generated
- 16 Oct 2026, 21:14
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy