The challenging world of in silico drug design: tools development and applications

Abstract

In the attempt to reduce time and costs of the drug discovery process, computational strategies have been looked as the possible solution. Despite still far from being the answer of all the problems, the use of computational approaches is now well established in the drug discovery pipeline. In the course of the years a lot of techniques have been developed and now are applied to several phases of drug discovery and development. In the present thesis is summarized the work conducted in three different projects carried out during my PhD, that allowed me to exploit different computational strategies. The goal of the main project was the optimization and the validation of the performance of a new drug discovery software, LiGen, result of the collaboration between the Italian pharmaceutical company Dompé, the Italian supercomputing center CINECA and our research group. LiGen was developed to perform protein surface analysis, molecular docking and de novo design; during this project we focused our attention mainly on the first two tools, LiGenPocket, aimed at the binding site analysis and structure-based pharmacophore definition, and LiGenDock, the molecular docking engine. Even if seldom used in computational chemistry, we decided to apply the Design of Experiments (DoE) methodology to optimize parameters controlling LiGenPocket and LiGenDock. At first we applied a fractional factorial design to screen the set of user-adjustable parameters to identify those having the largest influence on the accuracy of the results and then we optimize their values, to ensure the best performance in pose prediction and in virtual screening. Afterwards the results have been also compared with those obtained by two popular docking programs, namely Glide and AutoDock, for pose prediction, Glide and DOCK6 for Virtual Screening. The second project was the investigation of the binding mode of a series of compounds based on the 2-aminonicotinic 1-oxide scaffold and developed by our synthetic laboratory, to inhibit the 3-hydroxyanthranilic acid dioxygenase (3-HAO), an enzyme of the kynurenine pathway. 3-HAO is responsible for the production of the neurotoxic tryptophan metabolite quinolinic acid (QUIN); elevated brain levels of QUIN has been connected to several neurodegenerative diseases therefore 3-HAO inhibition may be a useful strategy for Huntington’s diseases and Alzheimer’s diseases among the others. To predict the most probable binding mode, compounds and the binding site have been characterized at quantum mechanical level, due to the presence of a catalytic iron atom in the binding site. Molecular docking was then used to predict the binding mode of the compounds and to investigate the effects of the substituents at the pyridine ring. The third project was related to the creation of a database of functional groups to screen chemical libraries, in order to reject or to flag these functionalities in libraries used for virtual screening purposes, in relation to their potential toxicity. The functional groups have been collected from different sources and have been classified according to the type of risk they may be related. This collection of compounds has been enriched also with compounds that have been identified as “frequent hitters”, indicating compounds often interfering in vitro assays, especially in HTS. The final database is therefore divided in three group, one collecting the intrinsically reactive moieties, one with functional groups susceptible to biotransformation into reactive metabolites and one containing substructures frequently identified as false positives in experimental tests.