Rising antimicrobial resistance demands new antibiotics, but discovering new therapeutics is often slow and laborious. Google Antigravity, an agentic development platform, aims to simplify and accelerate that process.
The platform provides an agentic science workbench that lets biologists rapidly discover, structurally evaluate, and prioritize promising targets. It compresses previously fragmented searches across multiple databases into a single session and enables researchers to scan millions of previously uncharacterized proteins in the UniProt database using ProtNLM, a model that predicts biological function from amino acid sequences.
Finding new antimicrobial peptides
Antimicrobial resistance is a major global health crisis; estimates project it could contribute to as many as 10 million deaths per year by 2050. To counter this threat, new therapeutics must be found urgently. Antimicrobial peptides (AMPs), particularly bacteriocins, are a promising alternative to traditional antibiotics because they naturally target fundamental vulnerabilities in competing bacteria — from cell envelopes to essential intracellular machinery — which makes it harder for pathogens to evolve resistance.
The exponential growth of genomic and metagenomic sequencing has populated global databases with billions of natural amino acid sequences. That pool has already yielded globally used bacteriocins like Nisin and recent genome‑mined discoveries such as Darobactin, Dynobactin, and Lariocidin. Yet many potentially useful proteins remain hidden in the unannotated “dark matter” of protein databases because they have no similarity to known proteins. The leading public resource for protein sequence and function information, UniProt, contains millions of sequences with unknown functionality.
To help bridge this gap, Google developed ProtNLM (Protein Natural Language Model), a deep learning model trained to read the evolutionary “grammar” of amino acid sequences. ProtNLM can predict biological function directly from raw protein sequences, enriching millions of previously unnamed UniProt entries with high‑confidence functional labels.
Collaborative intelligence in the workflow
Scientific discovery usually requires human intuition and iterative refinement. In Antigravity, a researcher and an AI agent collaborate: the researcher sets a focused biological objective — in the walkthrough, identifying novel AMPs from Streptococcus pseudopneumoniae — and through an interactive multi‑turn session directs which questions to ask and what evidence to gather. The agent performs cross‑database orchestration, structural analysis, and literature synthesis to support the investigation.
The candidate described in the technical walkthrough represents a biologically plausible lead supported by structural and genomic evidence and is now undergoing wet‑lab validation. The practical value of the platform is that it dramatically lowers the barrier for routine biological workflows, condensing days of fragmented database triage into a single collaborative session.
Although the demonstration focused on a narrow amino acid window in Streptococcus pseudopneumoniae, researchers can expand the conversation to include new filtering criteria, different sequence lengths, or other priority pathogens. By broadening searches from human commensals to environmental microbes, scientists can guide the agent across UniProt to discover novel AMP candidates at scale.
A shared 3D canvas for interactive collaboration
A protein’s 3D structure provides key insights into function, so examining a candidate’s predicted fold is essential. To enable an end‑to‑end workflow within one workspace, Antigravity includes the Biomolecular Structure Viewer, a 3D visualization tool embedded directly in the platform.
The researcher and agent operate over a shared visual state: as the agent retrieves structures from the AlphaFold Database or from the scientist’s own predictions, it can drive the viewer dynamically, highlighting functional regions to visually explain hypotheses. Crucially, the interaction is two‑way: researcher actions such as highlighting specific residues or pinning notes to a binding pocket are synchronized back to the agent. A researcher can 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 reason immediately over the exact structural context.
Accelerating scientific discovery
Historically, producing insights like those shown in the walkthrough required fragmented, specialized tools: custom scripts to query databases, manual exports to 3D modeling programs, disconnected genomic tools, and coordination among multiple experts to design wet‑lab assays. Because wet‑lab validation remains resource‑intensive, prioritizing the highest‑confidence leads early is essential.
Google Antigravity lets researchers navigate the vast biological search space efficiently while leveraging accurate protein function predictions from a domain‑specific model. By executing end‑to‑end scientific reasoning in a single environment, the platform enables scientists to spend more time evaluating high‑quality hypotheses, running laboratory experiments, and addressing research challenges.
Acknowledgements
Contributions to the project are acknowledged from: 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, Sebastian Nowozin.
