Google Antigravity adds cheminformatics skills, screens thousands of compounds in an hour
Google DeepMind launched cheminformatics skills for Antigravity on November 3, screening 568 compounds in about an hour and offering six free months to up to 10,000 scientists.
Google DeepMind announced on November 3 that it is launching a set of cheminformatics skills for Google Antigravity, covering literature search, molecular analysis, model building and new molecule design for drug discovery.
In a lung cancer demonstration, an AI agent retrieved thousands of compounds and ran matched molecular pair analysis in about an hour, proposing novel rigid macrocyclic molecules with synthesis routes, a process that previously took days. The tooling combines the planning, retrieval and scientific reasoning of large language models with deterministic chemistry tools such as RDKit, and connects to public databases including ChEMBL and PubChem. It identifies the same molecule appearing as an IUPAC name, a Markush structure or an SDF file across papers, patents and product catalogs, reducing data cleaning and format conversion effort.
The demo analyzed the lung cancer targeted drugs crizotinib and lorlatinib, retrieved and organized 568 compounds, identified a link between molecular rigidity and potency gains, trained a potency prediction model and designed several potentially better macrocycles with reaction mechanisms and synthesis suggestions.
Google emphasized that the output is a testable research hypothesis rather than a finished drug, since sparse data, inter-laboratory measurement error and limited dynamic range mean every candidate still requires synthesis and experimental validation. The feature is open to Antigravity users with enterprise access through Google Cloud accounts, and Google plans to offer six months of the highest usage tier free to up to 10,000 eligible scientists.