Google Research Builds AI Wet-Lab Agents, Finding Single-Digit Nanomolar Cancer Binders
Updated
Updated · Google Research · Oct 16
Google Research Builds AI Wet-Lab Agents, Finding Single-Digit Nanomolar Cancer Binders
3 articles · Updated · Google Research · Oct 16
Summary
Google Research said its AI agents can now plan, simulate and run wet-lab experiments through robotic systems, linking model reasoning directly to physical lab execution.
VERID, a live 3D digital twin of the lab, tracks robot geometry, reagent volumes and inventory so agents can catch impossible or flawed commands and refine protocols across runs.
Across 3 closed-loop aptamer design cycles, the agents lifted confirmed on-target hit rates to 33.5% for myoglobin and 42.1% for α-thrombin, learning from failures such as sticky off-target binders and disrupted folding.
In protein design against fibroblast activation protein alpha, Google said agent-guided epitope excision sped structure prediction 8.6-fold and helped produce single-digit nanomolar binders in a few experimental runs.
The work targets a broader goal of autonomous, human-supervised drug discovery, though Google said reliability, protocol verification and safeguards remain key hurdles.
Can Google's new AI agents bridge the gap between digital drug design and physical lab reality without making catastrophic errors?
How did an AI learn to outsmart deceptive biological traps to successfully design potent tumor-binding proteins from scratch?
Will autonomous wet-lab robots accelerate life-saving cancer treatments, or introduce unprecedented biosecurity risks if left unchecked?
From Eroom’s Law to Exponential Acceleration: How AI Wet-Lab Agents Are Transforming Drug Discovery and Biosecurity in 2026
Overview
By October 2026, Google DeepMind and Stanford University launched AI Co-Scientist, a Gemini-powered multi-agent system that autonomously generates, debates, and refines scientific hypotheses. This platform quickly identified new drug targets and a compound that blocked 91% of scarring in liver fibrosis models, while also accelerating data analysis in cellular rejuvenation research from months to days. To ensure safe and reliable lab automation, Google introduced the VERID digital twin framework, which prevents hardware errors before experiments run. These advances, combined with faster deep learning tools, have dramatically reduced drug development costs and timelines, reversing Eroom’s Law and expanding global AI-driven drug pipelines.