Modelling Cellular Permeability via Carrier Mediated Transport

Abstract

The relative importance of passive diffusion and carrier mediated transport processes to membrane permeability of drugs is a subject of current debate. Passive diffusion and carrier mediated transport are the two main methods by which drugs permeate the cell membrane. The permeability of molecules through membranes can have an impact on their absorption, distribution, metabolism and excretion (ADME) properties. It is therefore important to be able to predict the extent to which novel molecules can permeate the cell membrane. In vitro models of human intestinal absorption can be used to predict the likelihood of molecules permeating the human intestinal epithelium. Quantitative structure activity relationships (QSAR) techniques explain the relationship between molecular structure and cellular permeability. Current QSAR methods make use of physicochemical and structural property descriptors. These descriptors are able to predict the membrane permeability of molecules via passive diffusion rather than via membrane transporters. The aim of this study was to develop novel descriptors of carrier mediated transport that can be used in the development of QSAR models of permeability. The concept of metabolite likeness was investigated for its utility as a measure of the likelihood of molecules undergoing carrier mediated transport. This investigation found that approved drugs are generally more similar to human endogenous metabolites than molecules found in commercial databases. The use of a protein target prediction tool, PIDGIN, was also investigated. This study found that a relatively small number of membrane transporters that are expressed in caco-2 cells have models available in PIDGIN. New QSAR models of membrane permeability were developed using physicochemical and structural property descriptors and in combination with the novel descriptors of carrier mediated transport. Novel models for predicting drug efflux ratio were developed and perform well in validation tests. Comparisons of predictive performance between QSAR models generated from physicochemical property descriptors alone and in combination with ‘carrier-mediated transport descriptors’ were carried out. The general observation was that the novel descriptors of carrier mediated transport pursued did not significantly improve the predictive performance of models. However, some substructures from the MACCS keys list, which are relevant to protein binding, were found to be important determinants of caco-2 permeability of molecules and could potentially be used to identify molecules that may undergo active transport. The performance of logistic regression classification models of efflux ratio was 88%. Not many studies have developed QSAR models of efflux ratio. This is a relatively novel approach which could be useful in identifying, and thus help to avoid, potential substrates of efflux transporters in drug discovery.