Google DeepMind announced November 3 that its Antigravity agent now includes cheminformatics capabilities designed to accelerate drug discovery workflows. The new skills integrate RDKit, an open-source cheminformatics toolkit, with connections to the ChEMBL and PubChem databases.
According to DeepMind researchers Zoi Tsangalidou, Ruijie Zhang, Sukhdeep Singh and Michael V. LeVine, the additions allow the agent to search compound libraries, perform matched molecular pair analysis and propose synthetic routes. The team reports that in a demonstration using lung cancer treatment compounds, the agent completed a workflow including searching thousands of compounds and proposing “rigidified macrocycles with synthetic routes” in about an hour—a process they say typically requires multiple days.
The implementation separates tasks between the language model, which handles workflow orchestration and database navigation, and deterministic chemistry tools that execute molecular calculations. DeepMind describes this as addressing a persistent challenge in computational chemistry: molecules can be represented in multiple formats including IUPAC names, Markush structures in patents, or structure-data files in catalogs, and small errors can render results meaningless.
The cheminformatics skills are now available in Google Antigravity. DeepMind positions the additions as reducing time chemists spend on data preparation tasks.