New antibiotics are urgently needed as antimicrobial resistance rises, but the search for new therapeutics has long been a slow, labor-intensive process. Google its agentic development platform, Google Antigravity, can now make that process dramatically easier. In a technical walkthrough published on the Google DeepMind blog, researchers Pengfei Tian, Maxwell L. Bileschi, Zoi Tsangalidou, Sukhdeep Singh, and Lucy J. Colwell describe how biologists used the platform to uncover a novel antimicrobial peptide candidate with predicted activity against the pathogens involved in respiratory tract infections. The lead identified in the walkthrough is now undergoing wet-lab validation.

Antigravity functions as an agentic science workbench that empowers biologists to discover, structurally evaluate, and prioritize promising targets quickly, compressing what used to be a fragmented search across multiple databases into a single session. In practice, that means software agents plan and execute multi-step tasks — querying databases, retrieving structures, drafting protocols — while the scientist sets the strategy and weighs the evidence. The platform also helps researchers scan through millions of previously uncharacterized proteins in the UniProt database using ProtNLM, a model that predicts biological function from amino acid sequences.

The challenge of finding new antimicrobial peptides

Rising antimicrobial resistance is one of the most pressing global health crises of our time and, according to a UK government analysis, could result in an estimated 10 million annual deaths globally by 2050. Preventing that outcome, the authors argue, requires the urgent discovery of novel therapeutics.

Antimicrobial peptides (AMPs), particularly bacteriocins, offer a promising alternative to traditional antibiotics. They have naturally evolved to target fundamental vulnerabilities in competing bacteria, from cell envelopes to essential intracellular machinery, making it difficult for pathogens to develop resistance.

The exponential growth of genomic and metagenomic sequencing has populated global databases with billions of natural amino acid sequences. This pool of natural data has already produced widely used bacteriocins such as Nisin, alongside recent genome-mined breakthroughs including Darobactin, Dynobactin, and Lariocidin.

Many potential therapeutics, however, remain locked within the unannotated "dark matter" of protein databases, their functions uncharacterized because they bear little similarity to known proteins. UniProt, the world's leading public resource of protein sequence and functional information, holds millions of sequences with unknown functionality, transferred directly from genome sequencing projects.

To bridge this gap, Google developed ProtNLM (Protein Natural Language Model), a deep learning model trained to read the underlying evolutionary "grammar" of amino acid sequences. The model predicts biological function directly from raw protein sequences, enriching millions of previously unnamed UniProt entries with high-confidence functional labels. For unreviewed UniProtKB/TrEMBL entries, its annotation scope covers protein names, function comments, subcellular locations, GO terms, and keywords. That annotation layer is what makes uncharacterized sequences findable in biological terms, turning raw sequencing output into a searchable index of protein functions.

Human-AI collaboration in practice

Scientific discovery is rarely a single-step process; finding viable candidate proteins requires human intuition and iterative incorporation of new empirical evidence. In the walkthrough, the researcher and AI agent began with a focused biological goal: find and computationally validate novel AMPs in Streptococcus pseudopneumoniae. Across 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 workflow unfolded in several stages:

Navigating a vast search space. The agent queried the database and filtered millions of entries down to a handful of candidates worthy of attention, identifying a bacteriocin from the Lactococcin 972 family (UniProt ID: A0A0U0D591) with a length of 97 amino acids.

Validating the novelty of the finding. To verify the discovery was new, the researcher guided the agent to interrogate the candidate against curated knowledge in Swiss-Prot via the EMBL-EBI NCBI-BLAST service, confirming the peptide had no meaningful sequence homolog among manually reviewed proteins.

Reasoning over 3D protein structure. The researcher steered the analysis toward structural evidence. The agent retrieved the candidate's 3D model from the AlphaFold Protein Structure Database, queried it against experimentally determined structures to identify remote homologs, and superimposed the models to assess similarity. The structural evidence corroborated ProtNLM's prediction by showing the candidate's fold closely matches an experimentally determined bacteriocin, with the structure viewer in Antigravity allowing the analysis to be performed seamlessly in one place.

Reading the biological context around a gene. The agent looked beyond the molecule itself to its genomic neighborhood, reasoning over whether the surrounding machinery supported the AMP hypothesis. Inspecting the flanking genes revealed a telling pattern: the previously uncharacterized candidate sits upstream of a tightly coupled operon, just 6 base pairs ahead of an ABC transporter permease and its potential partner ATP-binding protein, alongside another annotated bacteriocin and an XRE transcriptional regulator. This provided strong contextual evidence that the candidate is part of a working bacteriocin system.

Building a mechanistic hypothesis from the literature. Reviewing existing literature, the agent proposed how the molecule might act and on which pathogens. In this case, existing evidence points to a mechanism of action targeting bacterial cell division, with a predicted potency spectrum spanning several drug-resistant pathogens.

Generating an experimental plan. Finally, the agent translated its in silico findings into an actionable bench protocol spanning protein expression, antimicrobial susceptibility testing, target engagement, and in vitro toxicity testing — bridging the gap between computational prediction and wet-lab experiment. That step is already in motion: the candidate, along with several others discovered through the same pipeline, is now in wet-lab validation, where experimental results will provide the ultimate test of the agent's analysis.

The candidate uncovered in the walkthrough represents a biologically plausible lead backed by structural and genomic evidence, but it still requires wet-lab validation. What makes this meaningful, according to the post, is how dramatically it lowers the barrier to everyday biological workflows, turning days of fragmented database triage into a single collaborative session. While the session focused on a narrow amino acid window in Streptococcus pseudopneumoniae, researchers can take the conversation in any direction — layering in new filtering criteria, exploring different sequence lengths, or targeting other priority pathogens. By broadening the search from human commensals to environmental microbes and beyond, scientists can guide the agent across UniProt to discover novel AMP candidates at scale. For anyone tracking whether AI-guided discovery can move from prediction to verified function, the results of these wet-lab validations are the next concrete signal to watch.

A shared 3D canvas for interactive collaboration

A protein's 3D structure can help explain its biological function, making it essential to examine a candidate's predicted fold in space. To provide this end-to-end workflow in a single workspace, Google is introducing the Biomolecular Structure Viewer, a 3D visualization tool embedded directly into Antigravity.

The researcher and agent work together over a shared visual state. As the agent retrieves structures from the AlphaFold Database or the scientist's own predictions, it can dynamically drive the viewer — for instance, highlighting key functional regions to visually explain its hypotheses. Crucially, this awareness flows both ways: actions such as highlighting specific residues or pinning notes directly onto a binding pocket are synchronized back to the agent. A researcher can simply mark a region of interest in 3D space and ask, for example, "How might this β-strand affect Lipid II binding?", allowing the agent to immediately reason over the exact structural context the scientist is looking at.

Accelerating scientific discovery

Historically, uncovering insights like those demonstrated in the workflow required fragmented and specialized software: writing custom scripts to query databases, manually exporting files to 3D modeling programs, managing disconnected genomic tools, and working with multiple experts to design wet-lab assays. Because wet-lab validation remains resource-intensive, narrowing the search space to only the highest-confidence leads upfront is critical. Google Antigravity allows researchers to navigate the vast biological search space efficiently while leveraging accurate protein function predictions from a domain-specific model, executing this end-to-end scientific reasoning in one place. That, the company says, lets scientists spend more time evaluating high-quality hypotheses, running experiments in the lab, and solving research challenges.

Acknowledgements

The post credits the following team members: 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.