Google Antigravity Finds Antimicrobial Peptide Lead in One Session

Google says researchers using its agentic development platform, Antigravity, have uncovered a novel antimicrobial peptide candidate with predicted activity against pathogens behind respiratory tract infections — and the lead is now undergoing wet-lab validation.

The claim comes from a technical walkthrough published by the Google team behind the platform. In it, biologists used Antigravity's agentic science workbench to discover, structurally evaluate and prioritize a promising target, compressing what the company describes as a fragmented search across multiple databases into a single session.

The stakes are hard to overstate. Antimicrobial resistance could cause an estimated 10 million annual deaths globally by 2050, according to figures cited by Google, making it one of the most pressing global health crises of our time. New antibiotics are urgently needed, but discovering them has traditionally been slow and labor-intensive — custom scripts, disconnected genomic tools, manual file exports and armies of specialists.

Antigravity is Google's attempt to collapse that pipeline into one workspace.

Why antimicrobial peptides matter

Antimicrobial peptides, particularly bacteriocins, offer a promising alternative to traditional antibiotics. Google's team explains why: these molecules have naturally evolved to target fundamental vulnerabilities in competing bacteria, from cell envelopes to essential intracellular machinery. That breadth of attack makes it difficult for pathogens to develop resistance.

The raw material for discovery already exists. Exponential growth in genomic and metagenomic sequencing has filled global databases with billions of natural amino acid sequences. That pool has already produced globally used bacteriocins like Nisin, alongside recent genome-mined breakthroughs such as Darobactin, Dynobactin and Lariocidin.

But much of it sits locked away. Google points to the "dark matter" of protein databases — millions of sequences in UniProt, the world's leading public resource of protein sequence and functional information, whose functions remain uncharacterized because they bear no similarity to known proteins. These entries were transferred directly from genome sequencing projects and never annotated.

What is ProtNLM?

To bridge that gap, Google developed ProtNLM, short for Protein Natural Language Model. It's a deep learning model trained to directly read the underlying evolutionary "grammar" of amino acid sequences, predicting biological function from raw protein sequences alone.

The model has already enriched millions of previously unnamed UniProt entries with high-confidence functional labels, according to the company. Inside Antigravity, researchers can use ProtNLM to scan those millions of uncharacterized proteins for candidates worth pursuing.

How did the discovery session work?

Scientific discovery is rarely a single-step process, Google's team notes. Finding viable candidate proteins requires human intuition and iteration to incorporate new empirical evidence. That's the design philosophy behind the collaboration.

Working in Antigravity, the researcher and AI agent began by setting a focused biological goal: find and computationally validate novel antimicrobial peptides in Streptococcus pseudopneumoniae. During an interactive, multi-turn session, the researcher steered the scientific strategy — deciding which questions to ask and what evidence to collect — while the agent handled the heavy lifting of cross-database orchestration, structural analysis and literature synthesis.

The result was a biologically plausible lead backed by structural and genomic evidence. Google is careful with its framing: the candidate still requires wet-lab validation, which is now underway.

What makes the result meaningful, the team argues, is how dramatically it lowers the barrier to everyday biological workflows — turning days of fragmented database triage into a single collaborative session.

What can researchers do next?

The walkthrough focused on a narrow amino acid window in one bacterium. But the platform doesn't stop there. Researchers can take the conversation in any direction, Google says:

• Layering in new filtering criteria
• Exploring different sequence lengths
• Targeting other priority pathogens
• Broadening the search from human commensals to environmental microbes and beyond

By widening the net across UniProt, scientists can guide the agent to discover novel antimicrobial peptide candidates at scale, according to the company.

What is the Biomolecular Structure Viewer?

Alongside the discovery walkthrough, Google introduced a new tool: the Biomolecular Structure Viewer, a 3D visualization instrument embedded directly into Antigravity.

A protein's 3D structure helps explain its biological function, and examining a candidate's predicted fold in space is a key step in validation. The viewer gives the researcher and agent a shared visual state to work over.

The agent retrieves structures from the AlphaFold Database or the scientist's own predictions, and can dynamically drive the viewer — highlighting key functional regions to visually explain its hypotheses, for example.

Crucially, that awareness flows both ways. When a researcher highlights specific residues or pins notes directly onto a binding pocket, those actions synchronize back to the agent. A scientist can mark a region of interest in 3D space and ask, "How might this β-strand affect Lipid II binding?" — and the agent immediately reasons over the exact structural context the scientist is looking at.

Why does this matter beyond one peptide?

Historically, this kind of workflow required fragmented and specialized software at every step: writing custom scripts to query databases, manually exporting files to 3D modelling programs, managing disconnected genomic tools and working with multiple experts to design the wet-lab assays.

Wet-lab validation remains resource-intensive, so narrowing the search space to only the highest-confidence leads upfront is critical, Google's team writes. Antigravity lets researchers navigate the vast biological search space efficiently while relying on accurate protein function predictions from a domain-specific model.

The company frames the payoff simply: executing end-to-end scientific reasoning in one place means scientists spend more time evaluating high-quality hypotheses, running experiments in the lab and solving research challenges.

The acknowledged team behind the work includes Maria Gabriela Cardoso, Victoria Johnston, Lexi Walker, Gena Hong, James King, Richard Green, Filip Radlinski, Artem Sobolev, Michael V. LeVine, Ruijie K. Zhang, Zachary Wu, Luca Schulz, Amy Danson, Tristan Oliver Kwan, Ryan Krueger, Oleg Kovalevskiy, Doug Fritz, Yunhan Xu and Sebastian Nowozin.

Whether the current lead survives wet-lab validation remains an open question — but the workflow that produced it is already reusable, and Google is betting that scaling this collaborative pattern across UniProt's millions of uncharacterized sequences will change how biologists hunt for therapeutics.

By Elena Vasquez