Google DeepMind's agent platform links more than 30 scientific databases to compress research workflows from days into minutes
Google DeepMind wants AI agents to do more than write code and summarize emails. Its Antigravity platform is now being pointed at drug discovery, where a single bad lead can burn months of lab time.
In an October 15, 2026 technical demonstration, the platform helped identify a new antimicrobial peptide candidate.
From millions of entries to one candidate
The demo centered on antimicrobial peptides, short chains of amino acids that can fight bacteria. Researchers used DeepMind’s ProtNLM model to sift through millions of unannotated UniProt entries.
UniProt is a large public protein database. “Unannotated” means many of those entries have no known function attached.
Out of that pile, the workflow surfaced a new peptide candidate. It shows promising activity against respiratory pathogens, specifically Streptococcus pseudopneumoniae.
The candidate was then checked in silico using AlphaFold, DeepMind’s protein structure prediction model. In silico simply means the testing happened on a computer rather than in a petri dish.
The process ended with a wet-lab experimental plan. That plan is currently in the validation phase, so the peptide has not yet proven itself in physical experiments.
What Antigravity actually does
At its core, Antigravity is a platform for AI agents working on life-science research. It integrates more than 30 scientific databases alongside DeepMind’s own models.
The pitch is speed. Exploring and analyzing biological data has historically taken days, according to the research findings. With Antigravity, researchers can expect to finish complex workflows in a matter of minutes.
The outputs are meant to be more than rough notes. The platform produces artifacts described as ready for publication, including 3D renderings and experimental plans.
Notably, scientists can generate those outputs without writing code.
Building on Science Skills
Antigravity did not appear from nowhere. It builds on the Science Skills suite, which Google introduced in May 2026.
Science Skills includes more than 11 tools covering proteomics and cheminformatics. Proteomics is the large-scale study of proteins, which makes it a natural partner for a peptide-hunting demo.
The suite aimed to streamline AI-assisted drug discovery across the biotech sector. Antigravity takes those tools and wraps them in an agent layer that can chain steps together on its own.
Google Cloud has also detailed a three-layer AI infrastructure built specifically for biotech drug discovery tasks.
