INTEGRATION OF QUANTUM-MECHANICAL AND MOLECULAR DOCKING APPROACHES IN GLOBAL METHODS FOR METABOLISM PREDICTION OF XENOBIOTICS

In silico methods represent a great opportunity to speed up some steps of the drug discovery process. Two in silico strategies that can be useful at different stages of the drug discovery process are presented here: in silico xenobiotic metabolism prediction and QSAR modeling. The former can be used to optimize the ADMET profile of a potential candidate, thereby eliminating undesirable compounds. The latter strategy can be used to identify new hits and for hit-to-lead optimization in the early stages of the drug discovery process. Overall, these two strategies can be seen as complementary and particularly useful during the preclinical phase. The lack of highly accurate metabolic resources hampers the development of machine learning algorithms for metabolic prediction. Recently, a manually curated database (MetaQSAR) has been proposed by our group. In addition, two in silico methods have been presented to predict both the occurrence of metabolic reactions (MetaClass) and the site of metabolism (MetaSpot). In both approaches, special attention has been given to the stereo-electronic descriptors calculated with MOPAC (PM7 Hamiltonian). Here, as a ligand-based approach, both methods have been extended using stereo-electronic descriptors calculated with density functional theory (DFT) to assess whether the use of features at a higher level of theory improves the overall performance. The CYP450s are known to be able to oxidize a large number of drugs and, together with other monooxygenases, pose a challenge to predicting metabolism using ligand-based approaches alone. Then, docking and molecular dynamic simulations were carried out using the bonded model within the MCPB.py approach, to efficiently parametrize the porphyrin ring. Then classification models were built using both ligand and structure-based information, provided by the molecular docking. In this way, experimental data on both cell viability and uncoupling/inhibitory activity of two niclosamide-derivatives datasets were correlated with molecular descriptors related to the proton shuttle mechanism to further advance the chemical synthesis of niclosamide-based compounds in collaboration with experimental groups.