Proof in Practice
Case Studies
Case Study 1
Leveraging Limited Data Using Federated AI Drug Design
Case Study 2
Saving years of med-chem costs
Case Study 3
Tailored ADME-T properties on demand
Qubigen’s Federated AI platform for Drug Design (FedAIDD) was tested used a limited/incomplete dataset for an oncogenic protein. AI designed a novel (potentially patentable) compound in 3 days with a similar binding mode to one in commercial development by a large pharma company, with favorable drug-like characteristics.
Using a client dataset targeting a membrane receptor, Qubigen’s FedAIDD replicated a blind holdback set of 27 molecules prioritized for further preclinical development using traditional SAR, plus 11 novel compounds with similar or better predicted properties. Novel compounds achieved sub-micromolar Kd in SPR binding assays.
Adding real-world ADMET data to the FedAIDD training pipeline for a transmembrane glycoprotein generated novel candidates with higher predicted solubility and lower predicted clearance - two critical properties for in vivo success. Higher solubility was verified for two AI-generated candidates.
Pipeline
