Google Antigravity gains chemistry skills for finding and designing molecules
Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .
Researchers can move more quickly from a body of publications to testable molecular candidates while leaving calculations to specialized chemistry tools.
Cheminformatics Skills connect a language model to RDKit, ChEMBL, and PubChem: the agent gathers compounds from the literature, cleans the data, compares molecular pairs, builds activity models, and proposes structures with synthesis routes. In a demonstration involving lung cancer drugs, it completed a cycle in one hour that went from thousands of compounds to candidate rigid macrocycles, whereas a conventional iteration takes several days. Deterministic chemistry tools perform the calculations while the model coordinates the work and explains its choices. The skills are available in Antigravity; proposed structures remain hypotheses and require synthesis and experimental validation.
Sources 1