Multidimensional (3D/4D-QSAR) probability-guided pharmacophore mapping: investigation of activity profile for a series of drug absorption promoters

Abstract

In the current study a hybrid approach that combines 3D and 4D-QSAR methods based on grid and neural (SOM) paradigms with automated variable elimination IVE-PLS procedure was examined to identify the pharmacophore pattern for cholic acid derivatives as potential drug absorption promoters. In particular, the outcome of multidimensional structure–activity modelling of the transdermal penetration effect (SKIN) and intestinal absorption enhancement (PAMPA) using the classical CoMFA and Hopfinger's cube formalisms has been compared with the neural CoMSA and SOM-4D-QSAR methodology for a set of cholic derivatives. The comparison of the corresponding statistic characteristics generally confirms the previously observed trends in pairs of q_cv²/q_test² values where 3D/4D SOM-based protocols with a fuzzy molecular representation for various training/test subset distributions outperforms the standard cubic 3D/4D procedures. A systematic model space inspection with splitting data collection into training/test subsets to monitor statistical performance in the effort for mapping of the probabilistic pharmacophore geometry was conducted using the stochastic SMV procedure. The iterative variable elimination procedure (IVE-PLS) represents a filter for specifying descriptors having potentially the highest individual weightings for the observed potency of cholic acid analogues as drug absorption promoters. A simplified visual inspection of pharmacophore sites gives the clear picture of regions that might be modified to modulate the compound potency. A pseudo-consensus 3D/4D-QSAR methodology was used to extract an average 3D pharmacophore hypothesis by exploration of the most densely populated training/test subpopulations to indicate the relevant factors contributing to the drug absorption potency of cholic acid derivatives.