Fragment based Drug Discovery; Design and Validation of a Fragment Library; Computer-based Fragment Screening and Fragment-to-Lead Expansion

In recent years, fragment screening has become a popular approach to identify new lead structures. Fragments are usually defined by the Astex ‘rule of three’ (RO3). Surface Plasmon Resonance (SPR), Nuclear Magnetic Resonance spectroscopy (NMR), biochemical assays and X-ray crystallography are effici...

Full description

Saved in:
Bibliographic Details
Main Author: Craan, Tobias Friedrich
Contributors: Klebe, Gerhard (Prof. Dr.) (Thesis advisor)
Format: Doctoral Thesis
Published: Philipps-Universität Marburg 2011
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!
Summary:In recent years, fragment screening has become a popular approach to identify new lead structures. Fragments are usually defined by the Astex ‘rule of three’ (RO3). Surface Plasmon Resonance (SPR), Nuclear Magnetic Resonance spectroscopy (NMR), biochemical assays and X-ray crystallography are efficient screening techniques to discover prospective fragments as binders. However, these methods need an assembled fragment library. We designed an in-house fragment library, starting from approx. 380,000 commercially available fragments. During library design, we modified the RO3 and we did no strict filtering of physico-chemical properties during fragment enumeration (e.g. twice the number of H-bond acceptors was allowed). The fragments were stepwise reduced to 4,000 compounds. The last step was a visual inspection of the candidates, which lead to a final fragment library of 364 fragments. To validate the quality of the library, we screened it against endothiapepsin. The biochemical screening suggested 55 hits, which were entered into a crystallographic screen. Eleven complex crystal structures were determined, pointing out the remarkably high hit rate of the designed library. HotspotsX is a program which predicts (based on knowledge-based potentials) the probability of a certain atom type at a certain position in the binding pocket of a target enzyme. The eleven crystal structures obtained before were used to validate the program HotspotsX. Due to chemical diversity and the different binding modes of the fragments observed for the library examples we obtained binding through aromatic- , H-bond donor- , acceptor- , doneptor- and hydrophobic interactions. The calculated HotspotsX maps coincide remarkably well with the crystallographically determined fragment positions inside the binding pocket. The program HotspotsX has also been validated with crystal structures of molecular probes like phenol, urea and methylurea. Crystal structures of these molecular probes were determined with different targets. Overall, the experimental hotspot analysis coincided well with the computed contour maps. Thus, the calculated maps by HotspotsX have an excellent predictive power. Based on the binding modes of the molecular probe phenol to the cAMP-dependent protein kinase A (PKA), we started a fragment growing approach. In the latter complex, three phenol molecules are bound. Two are occupying the ATP binding site and one is sitting on top of the glycine-rich loop (G-loop). A virtual screening, using the hinge binding phenol as constraint, suggested a phenol derivative for which a crystal structure could be determined. Starting from this hit, a hotspot analysis was performed. This analysis indicates that growth in the direction of the G-loop, placing an aromatic portion under the G-loop and an acceptor functionality capable to address Lys72 is desired. The first compound of this de novo design had an affinity of 70 µM. In the following first design cycle, we were able to enhance the affinity to 6.5 µM. In the second design cycle an additional amino function was introduced, which did not improve affinity dramatically, but enhanced ligand efficiency to 0.38. In the last cycle, a spacer of one and two methylene groups was introduced and the affinity could be increased to about 110 nM for a diastereomeric mixture of four compounds. The phenol-PKA complex provides a putative allosteric site of PKA. The G-loop in this structure is in a closed state which is stabilized by two H-bonds. This G-loop conformation is probably induced by the phenol molecule sitting on top of the G-loop. Therefore, several molecular dynamics (MD) studies were performed, lacking different phenol molecules, to get insights into the G-loop opening. The MD studies suggest that after removal of the phenol sitting on top of the G-loop some first side chain movements are initiated that can indicate the first steps of the G-loop opening cascade. In a different project, a virtual screening approach was used to find new inhibitors for aldose reductase. A pre-filtered subset of the ZINC database was used as ligand dataset. For the best hit, a series of five compounds was synthesized. Among them one compound displayed an inhibition of 920 nM. The available assays to detect fragment hits are currently not sufficient. The challenges are the low affinity of the fragments and their poor solubility. Therefore, the known thermal shift assay was applied and adapted to detect fragment hits. To validate the method, it was used to characterize variant mutations of EctD. Lastly, a modeling study was used to get ideas about possible binding modes of arachidonic acid derivatives in a K+ ion channel. One predominant binding pose could not be suggested. The study proposes, however, that one arachidonic acid molecule can occupy the inner pore cavity, which is consistent with experimental data.