The Pulse
Google Antigravity Finds a 97-Amino-Acid Antibiotic Lead
Google DeepMind says its Antigravity AI agent identified a 97-amino-acid antimicrobial peptide candidate in Streptococcus pseudopneumoniae. The candidate is predicted to act against respiratory pathogens, but it is still undergoing laborato
AI.info Team · Written with AI ·
A 97-amino-acid peptide is the leading candidate in a new antibiotic-discovery workflow demonstrated by Google DeepMind. Announced October 15, the candidate was found with Google Antigravity, the company’s agentic science workbench, and is predicted to act against pathogens involved in respiratory tract infections. It has not yet been shown to work in laboratory tests: Google says the candidate and several others from the same pipeline are undergoing wet-lab validation.
From millions of protein sequences to one lead
The search focused on antimicrobial peptides in Streptococcus pseudopneumoniae. Researchers set the biological goal, then worked with Antigravity’s AI agent to query databases, assess structures and review scientific literature. Google describes the workflow as a way to reduce a fragmented search across databases to a single session.
The agent identified a bacteriocin listed in UniProt as A0A0U0D591, a member of the Lactococcin 972 family. Google says its ProtNLM model helps search UniProt’s millions of previously uncharacterized proteins by predicting biological function from amino-acid sequences. The candidate had no meaningful sequence homolog in the manually reviewed Swiss-Prot collection, according to the company’s account of the analysis.
Sequence, structure and genomic clues
The team checked more than the sequence. Antigravity queried the candidate against curated protein records, retrieved its predicted structure from the AlphaFold Protein Structure Database and compared that model with experimentally determined structures. Google says the candidate’s fold closely matched an experimentally characterized bacteriocin, supporting ProtNLM’s functional prediction.
The agent also examined genes around the candidate. Google reports that it sits six base pairs upstream of a gene for an ABC transporter permease, near a possible partner ATP-binding protein, another annotated bacteriocin and an XRE transcriptional regulator. The company presents that arrangement as evidence consistent with the peptide belonging to a working bacteriocin system, not as proof that the molecule has been tested successfully.
Predicted activity still needs a lab result
By reviewing existing literature, the system proposed that the peptide may target bacterial cell division and could have activity against several drug-resistant pathogens. Those are predictions, rather than measured results from this candidate. Google says its planned bench work includes protein expression, antimicrobial susceptibility tests, target-engagement studies and in-vitro toxicity testing.
That distinction matters: a computationally plausible peptide is an early lead, not a treatment. The announcement does not report laboratory measurements showing that A0A0U0D591 kills bacteria, nor does it establish safety, dosing or effectiveness in animals or people. Wet-lab experiments are the next test of the agent’s analysis.
Antigravity puts the researcher in control
Google’s demonstration describes an interactive process rather than a fully autonomous discovery. A researcher chooses the target and steers the questions; the agent coordinates database searches, structural comparisons and literature synthesis. The workflow also introduces a Biomolecular Structure Viewer inside Antigravity, where researchers and the agent can inspect a protein model together and mark regions for further analysis.
Google frames the system as a way to reduce time spent moving between specialist databases, scripts and structural tools, leaving researchers with a shorter list of hypotheses to test. Whether that translates into useful antibiotics depends on the experiments now underway. For A0A0U0D591, those tests include expression, susceptibility, target engagement and toxicity.