Agentic cheminformatics with Google Antigravity
Reported by Google DeepMind1 source ↓
Google DeepMind says its new Cheminformatics Skills for Antigravity let an agent search compounds, analyze molecular changes and propose rigidified macrocycles with synthetic routes in about an hour, using a lung-cancer drug example.
Key points
- Google says the agent searched thousands of compounds and proposed macrocycles with synthetic routes in about an hour
- The demonstration used crizotinib and related compounds as a case study in molecular rigidification
- The skills use RDKit and connect to ChEMBL and PubChem
- Google reports that matched-pair analysis found cyclization generally increased potency and reduced rotatable bonds
Details
Google DeepMind’s demonstration follows a medicinal-chemistry workflow using crizotinib, a treatment for certain forms of non-small cell lung cancer, and related compounds. The lab says its Antigravity agent searched literature for compounds and associated data, then proposed rigidified macrocycles and synthetic routes in about an hour. The article does not provide a measured comparison of the proposed molecules’ potency or report experimental validation of the proposals.
The new skills combine language-model orchestration and scientific reasoning with deterministic chemistry calculations, according to Google. They use RDKit and connect to curated data from ChEMBL and PubChem; the agent also presents curated compounds and data in an interactive viewer.
Google says its matched molecular pair analysis compared compounds that differed in whether they were cyclized, and found that cyclization generally reduced rotatable bonds while increasing potency. The agent also built potency models with different features and architectures and tested them on a held-out set; the article gives no model scores in the available text.
Written from coverage as of
Images
- Google DeepMind
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh