Docking and QSAR comparative studies of polycyclic aromatic hydrocarbons and other procarcinogen interactions with cytochromes P450 1A1 and 1B1

Abstract

To obtain chemical clues on the process of bioactivation by cytochromes P450 1A1 and 1B1, some QSAR studies were carried out based on cellular experiments of the metabolic activation of polycyclic aromatic hydrocarbons and heterocyclic aromatic compounds by those enzymes. Firstly, the 3D structures of cytochromes 1A1 and 1B1 were built using homology modelling with a cytochrome 1A2 template. Using these structures, 32 ligands including heterocyclic aromatic compounds, polycyclic aromatic hydrocarbons and corresponding diols, were docked with LigandFit and CDOCKER algorithms. Binding mode analysis highlighted the importance of hydrophobic interactions and the hydrogen bonding network between cytochrome amino acids and docked molecules. Finally, for each enzyme, multilinear regression and artificial neural network QSAR models were developed and compared. These statistical models highlighted the importance of electronic, structural and energetic descriptors in metabolic activation process, and could be used for virtual screening of ligand databases. In the case of P450 1A1, the best model was obtained with artificial neural network analysis and gave an r ² of 0.66 and an external prediction of 0.73. Concerning P450 1B1, artificial neural network analysis gave a much more robust model, associated with an r ² value of 0.73 and an external prediction of 0.59.

Keywords:

• artificial neural network
• cytochrome P450 1A1
• cytochrome P450 1B1
• PAH
• QSAR