Drug Discovery Today  Volume 13, Numbers 7/8  April 2008 REVIEWS Computational models for prediction of interactions with ABC-transporters Gerhard F. Ecker 1 , Thomas Stockner 2 and Peter Chiba 3 1 Emerging Field Pharmacoinformatics, Department of Medicinal Chemistry, University of Vienna, Althanstrasse 14, 1090 Wien, Austria 2 Biogenetic and Natural Resources, Austrian Research Centers Seibersdorf, 2444 Seibersdorf, Austria 3 Institute of Medical Chemistry, Center for Physiology and Pathophysiology, Medical University of Vienna, Waehringerstrasse 10, 1090 Wien, Austria The polyspecific ligand recognition pattern of ATB-binding cassette (ABC)-transporters, combined with the limited knowledge on the molecular basis of their multispecificity, makes it difficult to apply traditional molecular modelling and quantitative structure–activity relationships (QSAR) methods for identification of new ligands. Recent advances relied mainly on pharmacophore modelling and machine learning methods. Structure-based design studies suffer from the lack of available protein structures at atomic resolution. The recently published protein homology models of P-glycoprotein structure, based on the high-resolution structure of the bacterial ABC-transporter of Sav1866, may open a new chapter for structure-based studies. Last, but not least, molecular dynamics simulations have already proved their high potential for structure–function modelling of ABC-transporter. Because of the recognition of several ABC-transporters as antitargets, algorithms for predicting substrate properties are of increasing interest. Introduction Transmembrane transporters are indispensably involved in the absorption, tissue distribution, excretion and toxicity, as well as pharmacokinetics and pharmacodynamics, of drugs. Members of the multidrug ATB-binding cassette (ABC) transporter subfamily have attracted particular interest, since they, in addition to their physiological role in tissue protection, actively extrude a large variety of therapeutically administered drugs from malignant cells and, thus, are responsible for multiple drug resistance in cancer patients [1]. Inhibition of the most intensively studied transpor- ters, ABCB1, ABCC1 and ABCG2, has been advocated as a mechan- ism for the restoration of drug sensitivity [2]. Additionally, there is increasing evidence that cholestatic forms of drug-induced liver damage result from a drug- or metabolite-mediated inhibition of hepatobiliary transporter systems, such as ABCB1, ABCB4, ABCG2, ABCG5 and ABCG8 [3]. Therefore, interaction with ABC-transpor- ters determines, to a large extent, the clinical usefulness, side effects and toxicity risks of drugs. Thus, detailed three-dimensional information on the molecular basis of drug-transporter interaction would have large potential value in assisting rational design of new drugs and establishing in silico models for the prediction of absorp- tion; distribution; metabolism; elimination; toxicity (ADMET) and safety problems. Most of the clinically relevant ABC-transporters, however, show a rather fuzzy and promiscuous pattern of ligand specificity. ABCB1, ABCC1 and ABCG2, the three key transporter involved in multiple drug resistance in tumour therapy, efflux a broad panel of structurally and functionally diverse compounds, which range from low molecular weight compounds such as cyclosporines, up to lipids [4]. This inherent promiscuity of ABC-transporters, accompanied by the limited knowledge on the molecular basis of this multispecificity renders traditional mole- cular modelling methods rather ineffective for generation of global predictive models. Nevertheless, there have been considerable modelling efforts to target promiscuous proteins, especially in the field of cytochromes [5] and the human ether-a-go-go-related gene (hERG) potassium channel [6], and the ABC-transporter field definitely can benefit from the experiences gained with these (anti)targets. Reviews  INFORMATICS Corresponding author: Ecker, G.F. (gerhard.f.ecker@univie.ac.at) 1359-6446/06/$ - see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2007.12.012 www.drugdiscoverytoday.com 311