Data-Driven Prediction of Risk in Drug Substance Starting Materials

Abstract

Regulatory approval for drug substance (DS) starting materials (SMs) represents a significant milestone in the progression toward approval of the proposed DS manufacturing process. To objectively predict viability of a proposed SM, a data-driven method has been developed that efficiently characterizes the risk associated with a potential SM designation. This method for prediction of risk in SMs (PRSM) is informed by an assessment of molecular and structural complexity, impurity risk, and propinquity to the DS. To develop the method, latent variable modeling was applied to identify molecular and synthetic route attributes that have correlated to historical agreement on proposed SMs by global regulatory agencies. As an outcome of the modeling approach, two metrics were empirically derived that identified high-risk SMs based on separate molecular complexity and impurity control factors. The utility of the classification system and associated method for PRSM have been tested and verified using both internal and publicly available external case studies.