Expanding the Toolbox for Computational Analysis in Rational Drug Discovery: Using Biomolecular Solvation to Predict Thermodynamic, Kinetic and Structural Properties of Protein-Ligand Complexes

Most biomolecular interactions occur in aqueous environment. Therefore, one must consider the interactions between proteins and water molecules when developing a drug molecule against a target protein. The study of these interactions is challenging using experimental techniques alone, therefore comp...

Full description

Saved in:
Bibliographic Details
Main Author: Hüfner, Tobias
Contributors: Klebe, Gerhard (Prof. Dr.) (Thesis advisor)
Format: Dissertation
Language:English
Published: Philipps-Universität Marburg 2019
Pharmazeutische Chemie
Subjects:
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Most biomolecular interactions occur in aqueous environment. Therefore, one must consider the interactions between proteins and water molecules when developing a drug molecule against a target protein. The study of these interactions is challenging using experimental techniques alone, therefore computer simulations are commonly used to study the molecular details of protein-water or ligand-water interactions. In the first study presented in this doctoral dissertation (Chapter 2), the development, parameterization and testing of an approach is presented that can be used to calculate the solvation contribution in protein-ligand binding thermodynamics. The approach uses an extensive amount of molecular dynamics trajectories in conjunction with GIST calculations in order to obtain models that can predict relative protein-ligand solvation thermodynamics. In order to validate the approach, the model system thrombin is investigated using a set of 53 ligands with experimentally characterized protein-ligand structures and ITC profiles. We found that the binding thermodynamics of 186 congeneric pairs of ligands can be accurately described using our solvation-based models. The relative free energy of binding for these 186 pairs can be calculated from the desolvation free energy of the ligand molecules alone. Furthermore, complete thermodynamic profiles for protein-ligand binding reactions (i.e. free energy, enthalpy and entropy of binding) are accurately predicted by incorporating GIST solvent data from the unbound ligand as well as the protein-ligand complex. In Chapter 3, the aforementioned approach is applied to develop a strategy that enables to equip drug molecules with a desired set of solvation thermodynamics properties. For this purpose, the thrombin ligands (same ligand series as in previous Chapter 2) and the corresponding GIST integrals are decomposed into smaller building block molecules. In the next step, the solvation thermodynamics for the building blocks in the ligand molecule as well as the solvation thermodynamics for the isolated building block in aqueous solution are calculated. We found greatly varying solvation thermodynamics for the different building blocks, demonstrating their potential to design ligands with a wide range of solvation characteristics. Also, we found that the building block decomposition of ligand molecules and the corresponding GIST integrals can be readily used to understand remote solvent structuring effects. These effects occur in the unbound ligand molecule and describe the enhanced solvent structuring on a building block in the ligand molecule due to the presence of another building block at a distal site of the ligand. Furthermore, we demonstrated that the fluorination of building blocks leads to an increased unfavorable desolvation free energy and thus disfavors binding for the presented dataset. The research presented in Chapter 2 and Chapter 3 was accomplished with the computer program Gips that was developed as part of this doctoral dissertation. In the following Chapter 4, the mechanism and time scale of desolvation is being analyzed for the protein-ligand dissociation reaction of trypsin and thrombin in complex with benzamidine and N amidinopiperidine. The analysis is carried out using umbrella sampling free energy calculations and LoCorA calculations. The LoCorA approach is a method for the analysis of residence times of water molecules on the surface of amino acids. It was found that water molecules reside approximately 1.3 ns in the binding pocket of thrombin, whereas in trypsin they are residing one order of magnitude shorter (0.3 ns). This difference is explained with special solvent channels that connect the interior of the binding pocket to bulk solvent environment. The solvent channels are present in thrombin but not in trypsin. Furthermore, the selectivity profiles of benzamidine and N amidinopiperidine are related to a solvent-mediated free energy barrier that is present in thrombin but not trypsin. Also due to the presence of the solvent channels, the water molecules show similar residence time for both complexes in the case of thrombin but differing residence times in the case of the two trypsin complexes. The LoCorA approach is implemented in the computer program LoCorA (same name as the approach itself), which was developed as part of this doctoral dissertation. In the course of this doctoral dissertation, further computational studies were carried out in combination with experimental ones. These can be found in chapter 5 of this dissertation. Each of these studies is preceded by a separate abstract and a statement concerning the author contribution.
Physical Description:203 Pages
DOI:https://doi.org/10.17192/z2019.0494