Google Antigravity Speeds Up Search for New Antimicrobial Peptides

Google DeepMind has unveiled a new workflow demonstrating how its Google Antigravity platform can dramatically speed up the hunt for antimicrobial peptides, tiny proteins that can kill bacteria and may help fight the growing crisis of drug-resistant infections.

How the platform works

Antimicrobial resistance is projected to cause up to 10 million deaths annually by 2050, making the search for new treatments an urgent priority. Antimicrobial peptides, especially a class called bacteriocins, are a promising avenue because bacteria find it harder to develop resistance to them than to conventional antibiotics.

The challenge is that global protein databases like UniProt contain millions of sequences with unknown functions — what researchers call the "dark matter" of protein data. To make sense of this vast pool, Google DeepMind developed ProtNLM, a deep learning model trained to predict what a protein does just by reading its amino acid sequence.

In the new walkthrough, a researcher and an AI agent working together in Antigravity set out to find novel antimicrobial peptides in the bacteria Streptococcus pseudopneumoniae, a common cause of respiratory tract infections. Over several interactive rounds, the researcher guided the strategy while the agent handled data searches, structural analysis, and literature review.

From database query to a candidate peptide

The agent quickly filtered millions of sequences down to a single promising bacteriocin candidate, identified by the UniProt entry A0A0U0D591. It is 97 amino acids long and belongs to the Lactococcin 972 family.

The team then verified that the peptide was genuinely novel by checking it against manually reviewed protein databases, confirming it has no close known relatives. Next, they examined its 3D shape using structures from the AlphaFold Protein Structure Database, finding that its fold closely matched an experimentally confirmed bacteriocin.

The agent also looked at the genes surrounding the candidate and found a telling pattern: it sat just six DNA units away from an ABC transporter system, a setup commonly associated with bacteriocin production. Finally, by reviewing existing research, the agent proposed that the peptide likely targets bacterial cell division and could be active against several drug-resistant pathogens.

What comes next

The lead peptide identified in this walkthrough is now moving into wet-lab validation, where it will be tested physically in a laboratory.

Google Antigravity frames this kind of work as a collaborative loop between human intuition and AI assistance. Instead of manually querying multiple databases and tools one by one, researchers can run a single session that orchestrates sequence searches, structural comparisons, and literature synthesis, compressing weeks of fragmented work into a much shorter time frame.

Google DeepMind published the full technical walkthrough on its blog, inviting other biologists to explore the platform for their own therapeutic discovery projects.

Key facts

• Google Antigravity is an agentic platform that lets researchers and AI agents work together in a single interactive session
• ProtNLM predicts the biological function of proteins directly from their amino acid sequences
• The platform helped identify a novel 97-amino-acid bacteriocin candidate against respiratory pathogens
• The lead peptide is now undergoing laboratory validation
• Antimicrobial resistance could cause 10 million deaths per year by 2050

Sources

1. Google Antigravity: Accelerating Antimicrobial Peptide Search — Google DeepMind

Every fact in this story comes from the source above.

Watch