Doris Alexandra Schütz

Title of the Doctoral Thesis: Kinetics in drug discovery - improved molecular understanding how to influence kinetic behavior of drug-like molecules

Publishing year: 2017

Tags: binding kinetics / QSKR, association rate / residence time / molecular dynamics / drug design


Abstract

Kinetics of small molecules has become an indisputable parameter in drug design during the recent years. Many research articles provided retrospective explanation on how the association rate constant and the dissociation rate constant of drugs to their targets could be optimized. Observations included rearrangement of the target protein, shielded hydrogen bonds or rearrangement of the water structure close to the binding site. However no prospective scheme how to trigger kinetic behaviour of the ligand could be retrieved. In this thesis we present a novel approach on how to influence association kinetics of drug-like molecules binding to the oncology target and chaperone Hsp90. In the framework of the IMI project “K4DD” (kinetics for drug discovery), in close collaboration with Merck KGaA Darmstadt, we could infer a rule for slowing down compounds associating to their drug target. In three different studies we showed how to rationally slow down molecules by chemically altering them. To our knowledge this is the first scheme to rationally slow down drugs associating to their targets. Being provided with a large dataset of Hsp90 inhibitors and a variety of X-ray structures of co- crystallized inhibitors, we were able to derive and validate a rule for kinetic behaviour of drugs. We employed correlation analysis of physicochemical descriptors, which resulted in significant contribution of polarity and furthermore desolvation free Energy, to impact association rates. We utilized steered Molecular Dynamics Simulations to describe the transition state, which poses the highest Energy barrier and therefore the rate-limiting step along the drug-binding pathway. Incorporating the results of the correlation analysis and the MD simulations, we concluded that introducing enthalpic desolvation penalties destabilizes the transition state, which slows down the on- rate of drug like compounds binding to the hydrophobic pocket of Hsp90. These findings are of substantial interest, as hydrophobic areas constitute an integral part of druggable sites in proteins. To our knowledge, this is the first validated scheme to prospectively optimize k on by chemical modification of drug-like molecules within a lead-like series. Furthermore this thesis deals with prediction of drug residence time employing Scaled Molecular Dynamics Simulations. Successful ranking of very diverse drug-like molecules was achieved by using the software Biki TM (BiKiTechnologies). Hsp90 inhibitors of different scaffolds, size and charge were simulated and the estimated residence time showed good correlation with SPR measurements obtained. This might be an approach to rank ligands and estimate their kinetic behaviour at an early stage in drug design. Classical Molecular Dynamics Simulations can due to hardware limitations, but also time-restrictions not be used for these kind of kinetic simulations, whereas Scaled MD demonstrates the potential of computing residence time in a fast and computationally less expensive manner. The residence time of the dataset predicted in this work, spans more than three orders of magnitude, which presents the largest range of residence time within a predicted datatset published.