Google DeepMind announced Cheminformatics Skills for Antigravity on November 3. The Science plugin combines RDKit chemistry tools with ChEMBL and PubChem connectors; language models coordinate the workflow while deterministic tools perform molecular calculations.

A vendor workflow demonstration

In Google’s example, the agent gathers compounds related to crizotinib and lorlatinib, runs matched molecular pair analysis, trains prediction models and proposes rigidified macrocycles with synthesis routes. Google says the workflow takes about an hour.

Predictions need experiments

The post explicitly warns that sparse data, assay noise and narrow measurement ranges limit predictions. Proposed compounds must be synthesized and validated experimentally. This is a workflow demonstration, not evidence that a new treatment works.

Google says the skills are available through the Antigravity app’s Science plugin, including an Enterprise route using a Business Account.