HOME- Bryn Mawr Conference
- Workshops & Training
- 2010 Oxford (Discovery)
- 2010 Oxford (ADMET)
- 2009 Oxford (Discovery)
- 2009 Oxford (ADMET)
- Bassan, A
- Cronin, M
- Hardy, B
- Helma, C
- Hopfinger, T
- Judson, P
- Leahy, D
- Madden, J
- Michielan, L
- Narayanan, D
- Myatt, G
- Obrezanova, O
- Thomas, S
- Zamora, I
- Poster Session
- Bursary Award
- 2008 Oxford
- 2006 Oxford
- Program
- Exhibition
- Registration
- Jobs
- Contact
- Support
- Schedule
|
|
|
|
|
|
|
Lisa Michielan obtained her undergraduate degree in pharmaceutical chemistry and technology at the University of Padova (Italy) in 2006. She is currently a Ph.D. student under the supervising Prof. Stefano Moro at the Molecular Modeling Section in the same university. Her research interests include the application of in silico ligand-based strategies, in particular Quantitative Structure Activity Relationships (QSARs), and chemometrics. Several 3D-QSAR models have been generated by combining the autocorrelation molecular electrostatic potential (autoMEP) vectors with linear and nonlinear approaches to predict both receptor binding affinity of human A2A adenosine receptor (hA2AR) antagonists and aqueous solvation free energy of organic compounds. Her results were described in a poster presentation at the 4th Joint Sheffield Conference on Chemoinformatics held in Sheffield in 2006. In the drug discovery process, the receptor subtype selectivity represents a crucial property to avoid limiting side-effects and efficacy problems of drug candidates. Recently, an alternative methodology, Support Vector Machine (SVM), has been considered to predict both potency profile and selectivity of human adenosine receptor antagonists. At the beginning of 2008 Lisa started a six-months collaboration at the Molecular Networks in Germany, under the supervising Prof. Johann Gasteiger. She focused on metabolism of xenobiotics by applying novel classification strategies for the prediction of the isoform specificity of cytochrome P450 substrates. According to the dispositions in the new REACH legislation, currently her studies are aimed at modeling both aquatic toxicity and modes of action of chemical compounds.
|
|
Classification Approaches for the Prediction of the Isoform Specificity of Cytochrome P450 Substrates and Modeling Aquatic Toxicity of Chemicals
Lisa Michielan (1), Lothar Terfloth (2), Johann Gasteiger (2) and Stefano Moro (1)
(1) Molecular Modeling Section, Dipartimento di Scienze Farmaceutiche, Università di Padova, via Marzolo 5, I-35131 Padova, Italy
(2) Molecular Networks GmbH, Henkestrasse 91, D-91052, Erlangen, Germany
Drug metabolism is a crucial pharmacokinetic step in the ADME properties to be investigated. The early efficient prediction of the metabolic fate of drug candidates represents a desired and needed tool to prevent experimentally time-consuming and expensive in vivo testing conditions in the drug discovery process. To date, several ligand-based modeling methods have been reported on the classification of cytochrome P450 (CYP450) substrates according to their route of metabolism. However, these strategies consider exclusive non-overlapping classes. We present the analysis of a collection of xenobiotics metabolized by five CYP450 enzymes by applying both multi- and single-label classification approaches. As it occurs in the real metabolism, we assumed that each compound can be metabolized by more CYP450 isoforms. If the xenobiotics are assigned simultaneously to multiple classes, the multi-label classification is the appropriate strategy. Cross-training with Support Vector Machine and Counter-propagation Neural Networks modeling methods in combination with several molecular descriptors have been utilized in the multi-label analysis. In the traditional single-label classification, the automatic variable selection was combined with Support Vector Machine (SVM). We have compared our models in the prediction of the isoform specificity of new chemical compounds.
Recently, in the field of toxicology, some important dispositions were issued in the REACH (Registration, Evaluation, Authorization of Chemical) law to improve the protection of human health and the environment, through a better identification and understanding of the chemical properties hazardous to safety. In this context, the aquatic toxicity is known as critical property affecting the toxicity profile of chemicals. The interest of our studies is to predict the presence and the mechanism of acute toxic effects of compounds by applying a classification modeling method. A well-defined threshold for acute toxicity has been established to divide our dataset in two classes by considering the environmental hazards set in the second revised edition of Globally Harmonized System of Classification and Labeling of Chemicals (GHS) by UNECE. We have developed a simple binary classifier in combination with several molecular descriptors as a useful strategy to assign chemical compounds to toxicologically-based classes. In more detail, three different types of molecular descriptors have been introduced in our SVM analysis: the autocorrelation molecular electrostatic potential (autoMEP) vectors, the Sterimol descriptors and the logP(o/w) property values. Moreover, this descriptor set has been extended by the introduction of further properties influencing the toxicodynamic profile. Then, a different SVM classifier has been derived to predict the modes of action (MOA) of toxic chemicals. The toxicokinetic and the toxicodynamic models may be applied in series for the prediction of both aquatic toxicity classes and modes of action of new compounds to assess their potential ecological risk and mechanism of toxicity. Some interesting results of our modeling experiments will be given.
|
|
|
|
|
|
|
|
|
