Google DeepMind says its agentic development platform, Google Antigravity, can shorten the early stages of antibiotic discovery. In a technical walkthrough, biologists used the workbench to search for antimicrobial peptides (AMPs) in Streptococcus pseudopneumoniae, evaluate the candidates' predicted structures, and settle on a lead with predicted activity against pathogens involved in respiratory tract infections. That candidate is now in wet-lab validation.
Antigravity brings several tools into one shared session. ProtNLM, a protein language model, assigns functional labels to raw amino acid sequences, helping researchers scan millions of unannotated entries in UniProt. A built-in Biomolecular Structure Viewer pulls structures from the AlphaFold Database and lets the agent and researcher interact with the same 3D view, so highlighting a residue or asking about a binding pocket can directly feed the next round of analysis.
The company argues that this workflow lowers a major barrier: previously, researchers had to move between custom database scripts, separate modelling tools, and multiple experts. By narrowing the search space before expensive wet-lab experiments, Antigravity could let scientists probe a wider range of microbes and uncharacterized protein "dark matter" for new bacteriocins, which are less prone to resistance than traditional antibiotics.