Agentic cheminformatics with Google Antigravity
deepmind ·
Google has introduced new Cheminformatics Skills for its Antigravity platform to accelerate drug discovery workflows that traditionally take multiple days. These skills enable AI agents to automate complex tasks such as data curation, molecular analysis, and candidate molecule proposal by integrating with tools like RDKit, ChEMBL, and PubChem. In a demonstration involving lung cancer treatments, the agent successfully analyzed thousands of compounds in just one hour, identifying trends between molecular rigidity and potency. The system performed matched molecular pair analysis and built predictive machine learning models to evaluate how cyclization affects drug efficacy. Based on these insights, the agent proposed novel rigidified macrocycles along with specific synthetic routes and reaction mechanisms. While the tool significantly reduces the time scientists spend on data wrangling, it emphasizes that computational designs remain hypotheses requiring experimental validation. This collaborative approach allows both computational and bench chemists to focus more on scientific reasoning than manual data processing.
deepmind