Virginie Y Martiny1, Pablo Carbonell2, Florent Chevillard3, Gautier Moroy1, Arnaud B Nicot4, Philippe Vayer5, Bruno O Villoutreix1, Maria A Miteva1. 1. Université Paris Diderot, Sorbonne Paris Cité, UMR-S 973 Inserm, Paris 75013, France, Inserm UMR-S 973, Molécules Thérapeutiques In Silico, Université Paris Diderot, Sorbonne Paris Cité, Paris 75013, France. 2. Research Programme on Biomedical Informatics (GRIB), Department of Experimental and Health Sciences, Universitat Pompeu Fabra, IMIM (Hospital del Mar Medical Research Institute), 08003 Barcelona, Spain. 3. Université Paris Diderot, Sorbonne Paris Cité, UMR-S 973 Inserm, Paris 75013, France. 4. Inserm U1064/ITUN, CHU, 44093 Nantes Cedex, France and. 5. BioInformatic Modelling Department, Technologie Servier, 45007 Orléans Cedex1, France.
Abstract
MOTIVATION: Cytochrome P450 (CYP) is a superfamily of enzymes responsible for the metabolism of drugs, xenobiotics and endogenous compounds. CYP2D6 metabolizes about 30% of drugs and predicting potential CYP2D6 inhibition is important in early-stage drug discovery. RESULTS: We developed an original in silico approach for the prediction of CYP2D6 inhibition combining the knowledge of the protein structure and its dynamic behavior in response to the binding of various ligands and machine learning modeling. This approach includes structural information for CYP2D6 based on the available crystal structures and molecular dynamic simulations (MD) that we performed to take into account conformational changes of the binding site. We performed modeling using three learning algorithms--support vector machine, RandomForest and NaiveBayesian--and we constructed combined models based on topological information of known CYP2D6 inhibitors and predicted binding energies computed by docking on both X-ray and MD protein conformations. In addition, we identified three MD-derived structures that are capable all together to better discriminate inhibitors and non-inhibitors compared with individual CYP2D6 conformations, thus ensuring complementary ligand profiles. Inhibition models based on classical molecular descriptors and predicted binding energies were able to predict CYP2D6 inhibition with an accuracy of 78% on the training set and 75% on the external validation set.
MOTIVATION:Cytochrome P450 (CYP) is a superfamily of enzymes responsible for the metabolism of drugs, xenobiotics and endogenous compounds. CYP2D6 metabolizes about 30% of drugs and predicting potential CYP2D6 inhibition is important in early-stage drug discovery. RESULTS: We developed an original in silico approach for the prediction of CYP2D6 inhibition combining the knowledge of the protein structure and its dynamic behavior in response to the binding of various ligands and machine learning modeling. This approach includes structural information for CYP2D6 based on the available crystal structures and molecular dynamic simulations (MD) that we performed to take into account conformational changes of the binding site. We performed modeling using three learning algorithms--support vector machine, RandomForest and NaiveBayesian--and we constructed combined models based on topological information of known CYP2D6 inhibitors and predicted binding energies computed by docking on both X-ray and MD protein conformations. In addition, we identified three MD-derived structures that are capable all together to better discriminate inhibitors and non-inhibitors compared with individual CYP2D6 conformations, thus ensuring complementary ligand profiles. Inhibition models based on classical molecular descriptors and predicted binding energies were able to predict CYP2D6 inhibition with an accuracy of 78% on the training set and 75% on the external validation set.
Authors: Johannes Hochleitner; Muhammad Akram; Martina Ueberall; Rohan A Davis; Birgit Waltenberger; Hermann Stuppner; Sonja Sturm; Florian Ueberall; Johanna M Gostner; Daniela Schuster Journal: Sci Rep Date: 2017-08-14 Impact factor: 4.379
Authors: Maxime Louet; Céline M Labbé; Charline Fagnen; Cassiano M Aono; Paula Homem-de-Mello; Bruno O Villoutreix; Maria A Miteva Journal: PLoS One Date: 2018-05-10 Impact factor: 3.240