(103e) Developing Quantitative Chemometric Models for Monitoring API Disproportionation Process with Raman Spectroscopy and X-Ray Diffraction

For API molecules that follow the “Spring and Parachute” model as a tablet goes through the stomach and gastrointestinal tract, the primary objective of tablet formulation is typically to mediate the sudden decrease of solubility by maximizing the hang-time. For development compound X, a formulation ingredient intended to manipulate the microenvironmental pH of the tablet triggered an out-of-proportion enhancement in performance. Investigation of the solid-state property of the freebase-salt continuum revealed a unique co-crystal which forms unexpectedly within the formulation. The co-crystal formation dramatically alters the path of the salt break-down, and subsequently contributes to a significant delay in the salt-to-freebase conversion.

X-ray diffraction and Raman spectroscopy were utilized to monitor the disproportionation process of the API in aqueous media in an in-vitro setting. Process samples (slurry) periodically taken from the vessel were filtered to enable X-ray analysis, while Raman provided continuous spectroscopic tracking of the API suspension in the system. Due to severe peak overlapping, the presence of multiple freebase polymorphs presented a tremendous hurdle to X-ray and Raman data analysis, limiting initial analysis to be (predominantly) qualitative in nature.

The quantitative aspect of the study yielded details of the mechanisms by which the API molecule trans-configure through multiple stages of solid state entities (salt, co-crystal, freebase polymorph 1, 2, 3...) as pH evolves in the GI tract. Mathematical algorithms were developed to address the challenges of quantitative analysis of spectroscopic datasets. A deconvolution algorithm based on random projection of powder X-ray diffractograms was developed to quantify the fraction of freebase polymorphs present in the process. A partial least square model was then built to relate the freebase fractions predicted from X-ray data to Raman spectroscopic data collected in situ. The partial least squares algorithm successfully modeled the covariance structure between Raman spectra and freebase fractions in the presence of severe peak overlapping. Together, these quantitative methods allowed for accurate and high-resolution experimental monitoring of the salt disproportionation process. More importantly, the mathematical methodologies presented in this talk provides a generalized framework that can be broadly applied to any system with a complex polymorph landscape involving spectroscopic data with overlapping peaks.