Google antigravity leverages agentic AI to accelerate drug discovery through cheminformatics

Google Antigravity has introduced new Cheminformatics Skills to significantly speed up drug discovery workflows, demonstrating the capability to search thousands of compounds and propose novel rigidified macrocycles with synthetic routes in just one hour, using lung cancer treatments as a case study. These skills address the complex challenge in cheminformatics where molecules can be represented in numerous formats, such as IUPAC names, Markush structures, or SDFs. The system integrates foundational tools like RDKit with curated data from ChEMBL and PubChem, allowing the language model to handle workflow orchestration while deterministic chemistry tools execute calculations. In a demonstration involving the TKI drugs crizotinib and lorlatinib, the agent successfully extracted and analyzed literature data, revealing that macrocycles tend to exhibit higher potency and fewer rotatable bonds. Furthermore, the system performed matched molecular pair analysis and built multiple predictive models, offering scientists the ability to iterate on design hypotheses, including proposing cyclization reactions with mechanistic diagrams. The platform is currently available via the Antigravity App's Science plugin, and Google is offering 10,000 scientists six months of free access to the highest usage tier. The development team includes contributors such as Filip Radlinski, Artem Sobolev, and Michael V. LeVine.