REVIEWS Drug Discovery Today  Volume 14, Numbers 7/8  April 2009 In silico platform for xenobiotics ADME-T pharmacological properties modeling and prediction. Part II: the body in a Hilbertian space Alexandre Jacob 1 , Jaturong Pratuangdejkul 2,3,4 , Se ´ bastien Buffet 1 , Jean-Marie Launay 2,3,4 and Philippe Manivet 2,3,4,5 1 Division of Structural Biology, BioQuanta, Paris 05, France 2 Franco-Thai Molecular Modeling Unit (FTMMU), Department of Microbiology, Faculty of Pharmacy, Mahidol University, Bangkok, Thailand 3 APHP, Ho ˆ pital Lariboisie `re, Service de Biochimie et de Biologie Mole ´ culaire, Unite ´ de Biologie clinique structurale, Paris 10, France 4 Universite ´ Paris Descartes, UFR des Sciences Pharmaceutiques et Biologiques, Paris 06, France 5 INSERM U829, Universite ´ d’Evry-Val d’Essone, Evry, France We have broken old surviving dogmas and concepts used in computational chemistry and created an efficient in silico ADME-T pharmacological properties modeling and prediction toolbox for any xenobiotic. With the help of an innovative and pragmatic approach combining various in silico techniques, like molecular modeling, quantum chemistry and in-house developed algorithms, the interactions between drugs and those enzymes, transporters and receptors involved in their biotransformation can be studied. ADME-T pharmacological parameters can then be predicted after in vitro and in vivo validations of in silico models. Introduction It is widely accepted that the accuracy of tools used in computa- tional chemistry and biology for the prediction of ADME-T proper- ties for any compound (xenobiotics) needs to be improved before they are used for modeling drugs’ toxicity and biotransformation. Unfortunately, no efficient solution has yet been proposed. Some attempts have been made to identify weaknesses in computational methods. For example, poor chemical structural diversity or dif- ferences in the chemical space observed between external and training molecule sets used in QSAR studies have been incrimi- nated in the lack of predictability of in silico ADME-T methods [1]. Also, people claim, in some publications, to have developed ‘high tech’ multidisciplinary platforms combining biological and infor- matics groups [2] with the intention of predicting ADME-T proper- ties for any chemical compounds with the help of ‘system biology’ algorithms [3]. In fact, obtaining efficient and accurate in silico ADME-T modeling and predictive tools is not so simple. It cannot be solved by simply adjusting the chemical space of molecules of the training and external test sets, as carried out by computational chemists with only a basic knowledge in biology, using inappropri- ate ‘magic algorithms’ that unduly increase the complexity of choosing the ideal in vitro experimental model. Existing solutions succeeded neither in reducing drugs’ clinical phase failures owing to toxicity discovered late in the process, nor in increasing opti- mization rate of pharmaceutical industry’s focused libraries by monitoring drugs’ ADME parameters. A real solution requires rethinking thoroughly the concept of ADME-T in silico, eliminat- ing old persisting dogmas in computational chemistry, once useful but that now prevent real progress in the field. We have shown in this issue that a ‘clear estimation of the accuracy of in silico methods’ [1] is not sufficient for their application within the REACH regulations (In silico platform for xenobiotics ADME-T pharmacological properties modeling and prediction: Part I. Beyond the reduction of animal model use). Indeed, it is quite important to train qualified professionals with both expertise in computational chemistry and biology (structural clinical biology) who will create in silico tools adapted to ADME-T prediction. We also showed the obvious necessity of acceptance by the molecular modelers of professional and legal responsibility on the results they provide if they want to be considered as evaluators. In this review, we will detail the architecture of a multidisciplinary in silico platform for ADME-T modeling and prediction, using stan- dardized methods, where total quality control management is an important part of the whole process ensuring the accuracy of in Reviews  INFORMATICS Corresponding author: Manivet, P. (philippe.manivet@lrb.aphp.fr) 406 www.drugdiscoverytoday.com 1359-6446/06/$ - see front matter ß 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2009.01.013