Google DeepMind’s AI platform accelerates discovery of antimicrobial peptide candidates

Researchers at Google DeepMind report using Google Antigravity, an agentic development platform, to accelerate the discovery of antimicrobial peptides (AMPs). The work targets rising antimicrobial resistance, which is projected to cause 10 million annual deaths globally by 2050. The team focused on bacteriocins, a class of AMPs that have evolved to target fundamental bacterial vulnerabilities, making resistance less likely to emerge quickly.

The researchers integrated ProtNLM, a deep learning model trained to predict biological function directly from amino acid sequences, enabling functional annotation of millions of previously uncharacterized proteins in UniProt. This allowed the team to enrich UniProt entries with high‑confidence labels for protein names, function comments, subcellular locations, and Gene Ontology terms.

Using Antigravity, the researchers conducted an interactive, multi‑turn session to set a focused goal: find and computationally validate novel AMPs in Streptococcus pseudopneumoniae. The agent orchestrated cross‑database queries, structural analyses, and literature synthesis while the biologist steered the strategy and decided which evidence to prioritize.

The agent filtered millions of entries down to a handful of candidates, including a bacteriocin with UniProt ID A0A0U0D591. That candidate has now entered wet‑lab validation for predicted activity against pathogens involved in respiratory tract infections.

The platform compresses what is typically a fragmented search across multiple databases into a single session, aiming to reduce the time required to identify promising therapeutic candidates.