Abstract Details

Poster 45: Computer Modelling of Synergistic Effects in Intestinal P-Glycoprotein Transport and Cytochrome P450 3A4 Metabolism

Alison Choy1, Andrew Howlett1, Johannes Kirchmair1, Robert Glen1
1Unilever Centre for Molecular Informatics, Department of Chemistry, Lensfield Road, Cambridge, CB2 1EW, UK
The oral bioavailability problems experienced by drugs in clinical trials are often due to the metabolism and efflux of xenobiotics by the body’s defence mechanism. Failures at clinical trials massively increase the cost of development of marketable drugs and should be avoided if possible. Drug efficacy problems can be caused by the metabolism and efflux of the compound by the body’s defence mechanism. The metabolic profile of a compound will also determine its exposure in vivo and the production and subsequent exposure of any metabolites formed. In silico tools that predict the metabolites formation and a compound’s ability to act as transporter substrate can both improve the efficiency of drug design and reduce the need for animal testing. Here, we focus on two detoxifying systems in the intestine, cytochrome P450 3A4 (CYP3A4) and P-glycoprotein (PGP), which are responsible for the metabolism and efflux of a wide variety of xenobiotics in the human intestine. The aim of this study is to produce in silico models to aid the understanding of how CYP3A4 and PGP contribute towards poor oral bioavailability and to integrate these models into an existing software tool for metabolism prediction. We hope the tool will allow easy estimation of the oral bioavailability of compounds at early stages of drug design so as to avoid costly failures later on in the development pipeline.
CYP3A4 is the most abundant and most promiscuous CYP isoform present in the human small intestine. It is also responsible for the metabolism of about 50% of all marketed drugs. Understanding the metabolism capacity of CYP3A4 will help explain a significant fraction of the oral bioavailability problems experienced by drugs and lead compounds. Aside from metabolism, the efflux of drug molecules also plays an important role in reducing the absorption of drug compounds. PGP is a transporter found in the apical membrane of intestinal epithelial cells and is an ATP-dependent drug efflux pump for a wide variety of xenobiotics and can cause multidrug resistance in cancer cells. It has been identified that these two detoxifying systems have significant overlaps in their substrate spectra, transcriptional regulation and expression patterns. These similarities have not yet been adequately taken into account by existing computational models and should not be overlooked. In this work, the possible interplay between the two detoxification systems was investigated.
2D descriptors were used to represent the input molecules as they are quicker to compute than 3D descriptors and will reduce the computational time required to process a large number of molecules – as would be the case in the early stages of a drug development program. Machine learning methods were used in an attempt to model the chemical space occupied by the compounds that are PGP substrates, CYP3A4 substrates and common substrates of both. It was found that there is a significant overlap between CYP3A4 and PGP substrates, although this is not surprising given the promiscuity of both detoxifying systems. The area-under-curve (AUC) of the best model for identifying CYP3A4 substrates against a background set was 82%. For PGP, the best model had an AUC of 81%. The best model for classifying molecules as substrates of neither, one or both systems produced an AUC of 81%.
As for the metabolic profile of xenobiotics, we use Metaprint2D as our tool for metabolism prediction. Metaprint2D predicts potential sites of metabolism (SOM) through a data-mining approach by identifying compounds containing similar fingerprints, resulting in the highlighting of positions predicted to be the most metabolically vulnerable. This approach is extended in Metaprint2D-React where metabolites most frequently observed for each of these SOMs are displayed and it also allows for the subsequent generations of daughter metabolites. In order to clearly present this information, the user can select which metabolites should be further processed for daughter metabolites generation. The enzyme responsible for the biotransformation giving rise to each metabolite is predicted by a protein target predictor and all metabolites can also be filtered by clogP.
We envisage that by integrating the models produced to predict CYP3A4 metabolism and PGP efflux into Metaprint2D-React, the software tool produced would allow the investigation of the metabolic spectra of input compounds, providing further insight into how metabolism and efflux are combined by the body in the process of reducing bioavailability of drugs and how the synergistic actions of certain detoxifying systems can bring about more extensive metabolism and extrusion of xenobiotic compounds.

Return to Programme