An amalgamation of LC-MS HDX and AI-based software for structure prediction of drug degradation products: An integrated step towards pharmaceutical development

Abstract

Structure identification of drug degradation products is one of the key components of any pharmaceutical product development. Knowledge-based in-silico tools are widely used to predict the probable degradation and excipient interaction products. However, multiple structures of the same masses are predicted through this software, which makes identification of the correct/exact structure of the degradation product difficult. In this study, the utilization of a simple yet powerful analytical tool, LC-MS HDX was explored for improving the predictive capability of the in-silico tool. For the same, 5 drugs were selected as representative of pharmaceutical development workflow and subjected to a stress degradation study as per ICH regulatory requirements. There were in total 55 degradation products, for which all possible structures were predicted through Zeneth^® software. The number of labile hydrogens in each of the 55 products was obtained through a simple LC-MS HDX method. While scrutinizing predicted structures with the number of labile hydrogens, many false positive predictions could be ruled out. Out of 55 degradation products, this approach has helped in improving the predictability of 45 products corresponding to ∼82% cases. It is simple to adopt for the industry by simply amalgamating analytical practices with AI prediction.

GRAPHICAL ABSTRACT

Keywords:

• Degradation impurity prediction
• Liquid Chromatography mass spectrometry
• Stress degradation
• Hydrogen-deuterium exchange