Data science replaces hazardous reagents in drug manufacturing
Genentech and Roche have developed a second-generation process for the intermediate GDC-6599, utilizing multivariate regression to eliminate toxic cyanide from the production line.
Dr. Ines Havel
Jul 3, 2026 · 1 min read
Safety and efficiency in pharmaceutical manufacturing often rely on the ability to replace legacy chemical routes with cleaner, more precise alternatives. In the development of GDC-6599, Genentech and Roche have successfully phased out a hazardous decakilogram-scale cyanation process. The second-generation route instead utilizes a five-step sequence featuring palladium-catalyzed arylation and rhodium-catalyzed hydroformylation, achieving a 38 percent overall yield.
This transition was made possible through the application of high-throughput experimentation and multivariate linear regression. By leveraging data science to select catalysts, the teams established the necessary stereochemical precision for the drug’s tetrahydrofuran core without the safety risks associated with cyanide reagents. The result is a more stable manufacturing protocol that provides tighter control over isomeric impurities while maintaining industrial scale.