Accelerating the search for antimicrobial peptides with Google Antigravity

Google Details Antigravity, An Agentic Platform for Scientific Discovery

Google has detailed its new agentic science workbench, Google Antigravity, demonstrating its capability to accelerate the search for novel antimicrobial peptides (AMPs). The platform was used to identify a new bacteriocin candidate with predicted activity against pathogens involved in respiratory tract infections, addressing the urgent need for new therapeutics to combat antimicrobial resistance. The system's primary function is to compress what is typically a fragmented, multi-day research process across numerous databases and software tools into a single, collaborative session between a biologist and an AI agent.

A Unified Technical Workflow

At the core of the platform is ProtNLM, a deep learning model developed by Google that predicts biological function directly from raw amino acid sequences. This allows the system to functionally annotate millions of previously uncharacterized proteins within the UniProt database. During a research session, the agent, guided by a scientist, performs a complex end-to-end workflow to identify and validate promising candidates.

• Database Search: The agent queries and filters millions of protein entries to isolate a small number of high-potential candidates based on the researcher's goals.
• Novelty Verification: It interrogates curated databases like Swiss-Prot via the NCBI-BLAST service to confirm that a candidate peptide is genuinely new and not a known protein.
• Structural Analysis: The platform retrieves 3D protein models from the AlphaFold Database and uses an integrated Biomolecular Structure Viewer to analyze folds, assess similarity to known structures, and facilitate collaborative annotation between the user and the agent.
• Genomic Context: The agent examines flanking genes to find contextual evidence, such as the presence of transporters or regulators, which supports the AMP hypothesis.
• Experimental Planning: Finally, it synthesizes its findings and relevant literature to propose an actionable protocol for wet-lab validation, which is currently underway for the identified lead candidate.

This unified approach allows researchers to navigate vast biological search spaces more efficiently, focusing their efforts on the highest-confidence leads. The integration of a shared 3D structure viewer is particularly notable, as it allows both the human researcher and the AI agent to reason over the same visual state, for instance, by letting a user highlight a binding pocket and ask the agent about its function directly.

Impact on Computational Biology

By orchestrating specialized tasks that previously required custom scripts and disconnected software, Google Antigravity lowers the technical barrier for conducting advanced biological research. It represents a shift from static tools to interactive, agent-driven scientific exploration. This could enable research institutions and pharmaceutical companies to scale their discovery pipelines, moving beyond AMPs to other therapeutic areas by allowing scientists to spend more time evaluating high-quality hypotheses rather than managing data and software.

Google Antigravity demonstrates a deliberate strategy to move beyond general-purpose AI assistants and create specialized, agentic workbenches that integrate directly into high-value R&D workflows, positioning Google's cloud and AI infrastructure as a core engine for scientific discovery.