Exploring the origin of phosphodiesterase inhibition via proteochemometric modeling

Abstract

The phosphodiesterase (PDE) superfamily, including all PDE families and subfamilies, are often implicated in diverse physiological disorders thereby making their selective inhibition of great necessity. Of the PDE4 family, the subfamilies of PDE4B and PDE4D have attracted attention due to their role in highly critical disorders such as asthma, acrodysostosis, cognition disorder and schizophrenia. Owing to their different levels of involvement in related disorders and within different subcellular compartments, the development of specific subfamily-selective compounds seems pertinent. Since achieving selectivity can be facilitated by considering the information of both compound and protein, thereby calling for proteochemometrics (PCM) to investigate the interaction space and selectivity of different chemical compounds towards different PDE4 isoforms. Several internal and external data sets were applied to validate the predictivity of the PCM model for interpolating on internal compounds as well as extrapolating on newly designed compounds. The Y-scrambling approach was applied to evaluate the possibility of chance correlation. Excellent values of 0.9973, 0.9037 and 0.9742 were observed for the training (R²), internal cross-validation (Q²) and external validation set (Q_ext²), respectively. Practical utilization of this information was demonstrated via the design of a few novel compounds whereby structural changes to the compound can exert effects on the selectivity against both PDE4B and PDE4D. Our model provided knowledge on the structural features of compounds in order to discriminate the binding of PDE4B and PDE4D, which is valuable for the promising design of selective inhibitors.