Google DeepMind adds cheminformatics skills to Antigravity, cutting days of drug-design work to about an hour
The skills, built on RDKit with ChEMBL and PubChem connectors, let an agent search thousands of compounds and propose rigidified macrocycles with synthesis routes.
Why it matters
In DeepMind's worked example on lung cancer drug data, an agent curated 568 compounds, ran matched molecular pair analysis, built potency models and proposed new 12-membered macrocycles with precursors and reactions in about an hour, a loop DeepMind says used to take multiple days. The post warns that sparse data and assay noise limit accuracy and that every design must be synthesised and tested. Google is offering 10,000 scientists its top Antigravity usage tier free for six months.
Key facts
- Built on RDKit with ChEMBL and PubChem connectors
- Worked example used 568 curated lung cancer compounds
- Proposed 12-membered macrocycles with synthesis routes in about an hour
- Designs still need synthesis and lab testing