IN SILICO APPROACHES IN DRUG DESIGN AND DEVELOPMENT: APPLICATIONS TO RATIONAL LIGAND DESIGN AND METABOLISM PREDICTION

In the last decades, the applications of computational methods in medicinal chemistry have experienced significant changes which have incredibly expanded their approaches, and more importantly their objectives. The overall aim of the present research project is to explore the different fields of the modelling studies by using well-known computational methods as well as different and innovative techniques. Indeed, computational methods traditionally consisted in ligand-based and the structure-based approaches substantially aimed at optimizing the ligand structure in terms of affinity, potency and selectivity. The studies concerning the muscarinic receptors in the present thesis applied these approaches for the rational design of novel improved bioactive molecules, interacting both in the orthosteric (e.g., 1,4-dioxane agonist) and in the allosteric sites. The research includes also the application of a novel method for target optimization, which consists in the generation of the so-called conformational chimeras to explore the flexibility of the modelled GPCR structures. In parallel, computational methods are finding successful applications in the research phase which precedes the ligand design and which is focused on a detailed validation and characterization of the biological target. A proper example of this kind of studies is given by the study regarding the purinergic receptors, which is aimed at the identification and characterization of potential allosteric binding pockets for the already reported inhibitors, exploiting also innovative approaches for binding site predictions (e.g., PELE, SPILLO-PBSS). Over time, computational applications felt a rich extension of their objectives and one of the clearest examples is represented by the ever increasing attempts to optimize the ADME/Tox profile of the novel compounds, so reducing the marked attrition in drug discovery caused by unsuitable pharmacokinetic profiles. Coherently, the first and main project of the present thesis regards the field of metabolism prediction and is founded on the meta-analysis and the corresponding database called MetaSar, manually collected from the recent specialized literature. This ongoing extended project includes different studies which are overall aimed at developing a comprehensive method for metabolism prediction. In detail, this Thesis reports an interesting application of the database which exploits an innovative predictive technique, the Proteochemometric modelling (PCM). This approach is indeed at the forefront of the latest modelling techniques, as it perfectly fits the growing request of new solutions to deal with the incredibly huge amount of data recently produced by the “omics” disciplines. In this context, MetaSar represents an alternative and still appropriate source of data for PCM studies, which also enables the extension of its fields of application to a new avenue, such as the prediction of metabolism biotransformation. In the present thesis, we present the first example of these applications, which involves the building of a classification model for the prediction of the glucuronidation reaction. The field of glucuronidation reactions is exhaustively explored also through an homology modelling study aimed at defining the complete three-dimensional structure of the enzyme UGT2B7, the main isoform of glucuronidation enzymes in humans, in complex with the cofactor UDPGA and a typical substrate, such as Naproxen. The paths of the substrate entering to the binding site and the egress of the product have been investigated by performing Steered Molecular Dynamics (SMD) simulations, which were also useful to gain deeper insights regarding the full mechanism of action and the movements of the cofactor.