Modélisation de molécules organiques hétérocycliques biologiquement actives par des méthodes QSAR/QSPR. Recherche de nouveaux médicaments
Samir Chtita · HAL (Le Centre pour la Communication Scientifique Directe) · 2017
The experiment is a straight way to obtain the activity/propriety data of organic compounds. Such experiment may be deficient in terms of requiring various sample organs, costing much money, taking much time as well different measured values used by different researchers. Consequently, it would be almost impossible for these experiments to provide the activity values of all organic compounds. Hence, it is crucial to use theoretical methods to make up for the disadvantages of the experiment and to predict the exact data of compounds.The significant development of computer science as well as the theoretical quantum of chemical studies enables researchers to get more precise physicochemical and quantum parameters of compounds in a shorter time.The structural parameters along with the introduction of the quantitative structure activity/propriety relationship QSAR/QSPR methods can increase the interpretability and predict the activity/propriety of new organic compounds. This is used to examine the relationship between molecular descriptors of a set of compounds and their biological activity or physicochemical propriety. Therefore, the author will elaborate, in this dissertation, on the antileishmanial activity, DNA drug binding proprieties, anticancer activity and antagonistic activity against the NMDA receptor of some organic heterocyclic, such as Acridine, Isatin, and Dizocilpine (MK801) derivatives. Quantum chemical calculation use density functional theory, DFT, with Becke’s three-parameters hybrid function, B3, and Lee-Young-Parr, LYP, exchange correlation functional methods. These methods are performed on the studied compounds and are used to calculate the electronic and quantum chemical parameters. A variety of molecular descriptors are computed with Gaussian, ACD/ChemSketch, Marvin Sketch, and ChemOffice programs. The datasets are subject to multivariate statistical analyses, i.e. principal components analysis PCA, multiple linear regression MLR, multiple nonlinear regression MNLR, partial least squares PLS, and artificial neural network ANN. Both obtained linear and nonlinear models are proposed and validated according to the principles that are set up by the Organization for Economic Co-operation and Development (OECD). The applicability domain of models is investigated using William’s plot to detect outlier and outside compounds. To successfully apply the developed models in order to predict new compounds, rigorous validations have been used in this direction. The effects of different descriptors in the activities/proprieties are described and used to examine and form new compounds with larger effect values (activity or propriety).