posted on 2024-02-19, 20:37authored byRachel Hemingway, Steven W. Baertschi, David Benstead, John M. Campbell, Maggie Coombs, Zongyun Huang, Ryan Khalaf, Martin A. Ott, David J. Ponting, Darren L. Reid, Neil G. Stevenson, Samuel J. Webb, Angela White, Todd Zelesky
Zeneth, a software
application for the prediction of chemical degradation
of small organic molecules, incorporates a knowledge base of rules
to predict degradation pathways. In addition, the knowledge base
contains property predictors that modulate the predicted likelihood
of a given degradation product. In this study, a C–H bond dissociation
energy (C–H BDE) predictor, which has been integrated into
the software, was utilized. To determine this software’s predictive
capabilities [using its knowledge base (2020.1.0 KB)], experimentally
derived degradation profiles for 25 drug substances were compared
to Zeneth predictions. These degradation profiles were derived from
forced degradation studies, including accelerated and long-term stability
studies, aligned with International Council for Harmonisation (ICH)
guidelines. In addition, two case studies highlighting how prediction
data can be utilized to confirm experimental data or assist with the
identification of unknown degradation products have been presented.
The specificity of prediction results was evaluated; transformation
types that often predict degradation products not observed experimentally
were identified, and an assessment of the causes is presented. The
sensitivity for the study group was also evaluated using a historic
knowledge base (2012.2.0 KB), enabling an assessment of how the predictive
capabilities have improved over this period; the comparison demonstrated
a 40% increase in sensitivity. This study has demonstrated that the
ongoing expansion and optimization of this in silico tools knowledge
base continues to result in improvements in its predictive capability
and its ability to impart insight into the drug degradation knowledge
space to aid pharmaceutical development.