Google has introduced a suite of Cheminformatics Skills for its Antigravity platform, designed to help chemists work with molecular data and carry out drug-discovery workflows. The skills aim to reduce time spent reconciling chemical information scattered across papers, patents and vendor catalogs, where the same compound may appear in different formats.
The system divides tasks between a language model, which coordinates workflows, navigates databases and applies scientific reasoning, and deterministic chemistry software, which performs molecular calculations. Google says the skills use RDKit, an open-source cheminformatics toolkit, and connect to curated resources including ChEMBL and PubChem.
The announcement includes a technical example focused on molecular rigidification, a medicinal-chemistry strategy that can improve a drug candidate’s potency. Molecules with rotatable bonds can adopt many shapes; adding a bond to constrain their movement may make it easier for them to retain the shape needed to bind a target.
Google uses crizotinib and lorlatinib to illustrate the approach. Crizotinib has five major rotatable bonds, while lorlatinib, a related third-generation tyrosine kinase inhibitor, is cyclized at one end to reduce its effective number of rotatable bonds to zero. The post presents this change as an example of rigidification and its potential to improve potency.
Sources: Google DeepMind Blog.