Google DeepMind uses Antigravity to search for antimicrobial peptides and adds a 3D protein structure viewer

發生什麼事

In this technical demo, screening began with millions of protein sequences in UniProt whose function is unknown. Google's ProtNLM model inferred their functions from the sequences, and the agent used those labels to narrow the field (Google DeepMind blog).

Literature reviews, structural comparisons and lookups across databases were the agent's job, ending in a written experiment plan, while researchers still chose where the research went. Antigravity has also gained a Biomolecular Structure Viewer. Markings that researchers make on a 3D structure are passed to the agent, and the agent in turn can drive the viewer to show its reasoning. The post doesn't state who will be able to use the viewer (Google DeepMind blog).