Development and prospective application of chemoinformatic tools to explore new ligand chemistry and protein biology

Drug discovery and design is a tedious and expensive process whose small chances of success necessitates the development of novel chemoinformatic approaches and concepts. Their common goal is the efficient and robust identification of promising chemical matter and the reliable prediction of its properties. Computer-aided drug discovery and design (CADDD) and its multifarious installments throughout the different phases of the drug discovery pipeline contribute significantly to the expansion of the hits, the understanding of their structure-activity relationship and their rational diversification. They alleviate the development’s costs and its time-demand thus support the search for the needle in the haystack – a potent hit. The HTS-driven brute-force nature of current and of the decades’ past discovery and design strategies compelled researchers to develop ideas and algorithms in order to interfere with the pipeline and prevent its frequent failures. In the introduction, I describe the drug discovery and design pipeline and point out interfaces where CADDD contributes to its success. In Part 1 of this thesis, I present a novel methodology that supports the early-stage hit discovery processes through a fragment-based reduced graph similarity approach (RedFrag). It is a chimeric algorithm that combines fingerprint-based similarity calculation with scaffold-hopping-enabling graph isomorphism. We thoroughly investigated its performance retro- and prospectively. It uses a new type of reduced graph that does not suffer from information loss during its construction and bypasses the necessity of feature definitions. Built upon chemical epitopes resulting from molecule fragmentation, the reduced graph embodies physico-chemical and 2D-structural properties of a molecule. Reduced graphs are compared with a continuous-similarity-distance-driven maximal common subgraph algorithm, which calculates similarity at the fragmental and topological levels. The second chapter, Part 2, is dedicated to PrenDB: A digital compendium of the reaction space of prenyltransferases of the dimethylallyltryptophan synthase (DMATS) superfamily. Their catalytical transformations represent a major skeletal diversification step in the biosynthesis of secondary metabolites including the indole alkaloids. DMATS enzymes thus contribute significantly to the biological and pharmacological diversity of small molecule metabolites. The attachment of the prenyl donor to lead- or drug-like molecules renders the prenyltransferases useful in the access of chemical space that is difficult to reach by conventional synthesis. In PrenDB, we collected the substrates, enzymes and products. We then used a newly developed algorithm based on molecular fragmentation to automatically extract reactive chemical epitopes. The analysis of the collected data sheds light on the thus far explored substrate space of DMATS enzymes. We supplemented the browsable database with algorithmic prediction routines in order to assess the prenylability of novel compounds and did so for a set of 38 molecules. In a case study, Part 3, we investigated the regioselectivity of five prenyltransferases in the presence of unnatural prenyl donors. Detailed biochemical investigations revealed the acceptance of these dimethylallyl pyrophosphate (DMAPP) analogs by all tested enzymes with different relative activities and regioselectivities. In order to understand the activity profiles and their differences on a molecular level we investigated the interaction within the enzyme-prenyl donor-substrate system with molecular dynamics. Our experiments show that the reactivity of a prenyl donor strongly correlates with the distance of its electrophilic, reactive atom and the nucleophilic center of the substrate molecule. It renders the first step towards a better mechanistic understanding of the reactivity of prenyltransferases and expands significantly the potential usage and rational design of tryptophan prenylating enzymes as biocatalysts for Friedel–Crafts alkylation. Lastly, in Part 4, we present the synergistic potential of combined ligand- and structure-based drug discovery methodologies applied to the β2-adrenergic receptor (β2AR). The β2AR is a G protein-coupled receptor (GPCR) and a well-explored target. By the joint application of fingerprint-based similarity, substructure-based searches and docking we discovered 13 ligands – ten of which were novel – of this particular GPCR. Of note, two of the molecules used as starting points for the similarity and substructure searches distinguish themselves from other β2AR antagonists by their unique scaffold. Thus, the usage of a multistep hierarchical or parallel screening approach enabled us to use these unique structural features and discover novel chemical matter beyond the bounds of the ligand space known so far and emphasize the intrinsic complementarity of ligand- and structure-based approaches. The molecules described in this work allow us to explore the ligand space around the previously reported molecules in greater detail, leading to insights into their structure-activity relationship. In addition, we also characterized our hits with experimental binding and selectivity data and discussed it based on their putative binding modes derived by docking