Google DeepMind Releases Cheminformatics Skills for Antigravity to Accelerate Drug Discovery

On November 3, 2026, Google DeepMind announced the release of "Cheminformatics Skills" for its Google Antigravity agent platform. This update aims to assist computational and experimental chemists in handling complex drug discovery workflows, including extracting data from literature, performing molecular analysis, and proposing novel candidate compounds.

The core of this skill suite lies in its division of labor: the underlying large language model handles workflow orchestration, database navigation, and scientific reasoning, while deterministic chemistry tools execute specific molecular calculations. Built on the open-source toolkit RDKit and connected to authoritative databases like ChEMBL and PubChem, these skills allow AI agents to not only understand natural language instructions but also invoke professional tools for precise chemical operations, avoiding potential hallucinations or calculation errors inherent in relying solely on language models.

DeepMind demonstrated a case study focused on non-small cell lung cancer treatments. Traditionally, curating data from literature, training predictive models, and proposing new molecules could take several days. Using Antigravity's Cheminformatics Skills, an AI agent completed the following tasks in approximately one hour: searching through thousands of compounds, filtering out irrelevant molecules, performing matched molecular pair analysis (MMPA) to quantify the impact of structural changes on activity, and proposing novel rigidified macrocycles with synthetic routes based on rigidity strategies. For instance, the agent analyzed crizotinib and lorlatinib, identifying how the latter improves binding affinity by reducing rotatable bonds, and designed new rigid macrocyclic molecules accordingly.