Computational Studies of Macrocycles and Molecular Modeling of Hepatitis C Virus NS3 Protease Inhibitors

Abstract: Computational tools are utilized in the drug discovery process to discover, design, and optimize new therapeutics. One important approach is structure-based drug design which relies on knowledge about the 3D structure of the biological target. The first part of this work focuses on applying structure-based drug design for binding mode prediction of HCV NS3 protease inhibitors. The NS3 protease is a challenging target from a computational perspective as it contains an extended binding site. Binding mode predictions were performed for various classes of new acyclic and macrocyclic HCV NS3 protease inhibitors and was used in the design of new inhibitors. None of the synthetized inhibitors have been co-crystallized yet, which has made the evaluation of the suggested binding mode predictions challenging.Macrocycles are an interesting compound class in drug discovery due to their unique structural architecture, which can enable access to new chemical space. Macrocycles can successfully modulate difficult therapeutic targets, as exemplified in the development of protease inhibitors. Furthermore they can improve drug-like properties, such as cell permeability and bioavailability. The second part of this thesis focuses on macrocycles from a computational point of view. A data set of 47 clinically relevant macrocycles was compiled and used in these studies. First, two different docking protocols rigid docking of pre-generated conformers and flexible docking in Glide were evaluated and compared. The results showed that flexible docking in Glide was sufficient for docking of macrocycles with respect to accuracy and speed.The aim of the second study was to evaluate and compare the performance of the more general conformational analysis tools, MCMM and MTLMOD, with the recently developed macrocycle-specialized conformational sampling tools, Prime-MCS and MMBS. In most cases, the general conformational analysis tools (with enhanced parameter settings) performed equally well as compared to the macrocycle-specialized conformational sampling techniques. However, MMBS was superior at locating the global energy minimum conformation.Finally, calculation of the conformational energy penalty of protein-bound macrocycles was performed. The macrocycle data set was complemented with linear analogues that are similar either with respect to physicochemical properties or 2D fingerprints. The conformational energy penalties of these linear analogues were calculated and compared to the conformational energy penalties of the macrocycles. The complete data set of macrocycles and non-macrocycles in this study differ from previously published work addressing conformational energy penalties, since it covers a more extended area of chemical space. Furthermore, there was a weak correlation between the calculated conformational energy penalties and the flexibility of the structures.