Computational tools for the development of screening studies

This PhD thesis was focused on the exploration and study of different computational strategies that are employed in computer-aided drug design and virtual screening (VS) campaigns aimed at the identification of novel ligands of different protein targets. The aim of the thesis was to validate the reliability of these procedures and to combine them in order to develop efficient virtual screening protocols. Ligand-based, pharmacophore-based and receptor-based approaches have been studied and combined together to exploit the advantages of all the different screening methods for the identification of new ligand hits. In particular, the first chapter of this manuscript reports the development of a VS protocol combining the ligand-based filtering of a commercial database with a docking and rescoring approach followed by molecular dynamic (MD) simulations that allowed the identification of new Fyn kinase inhibitors. In the second chapter, the development and validation of a consensus-docking approach is described. This procedure, which combines the results of several docking methods to improve docking accuracy and reliability, has been assessed from both the qualitative and quantitative point of view. Based on the promising results obtained, the consensus docking approach has been then experimentally validated applying the procedure in a VS study that led to the discovery of new non-covalent FAAH inhibitors (chapter 3). Chapter 4 describes how a new inhibitor of STAT3 dimerization has been discovered through the combination of a receptor-based pharmacophore search with different docking studies and MD simulations analyses, while the last chapter is dedicated to the study of conformational sampling through a compared assessment of two different conformer generators, which allowed to derive some guidelines for the generation of suitable conformational ensembles of small molecules.