Computational Methods for the Modulation of Protein-Protein Interactions

Abstract

During the last decades, drug discovery development has made considerable progress. However, annual numbers of released drugs for novel targets have been decreasing concomitantly. Limited success rates of combinatorial chemistry and high-throughput screening, as well as availability of feasible targets are some reasons for this problem. A strategy to overcome it is exploration of novel target classes in order to expand the druggable space. An example are protein-protein interactions (PPIs) that can be inhibited or stabilized. Inhibition aims at developing binders for one protein to prevent complex formation. However, known PPI inhibitors differ significantly from conventional drugs and current active site-biased compound libraries are probably inappropriate to discover them. The design of novel screening libraries is thus very important. PPI stabilization aims at developing molecules that bind to a protein complex to increase its stability like a molecular glue. In contrast to inhibition, it is rather unexplored but ground-breaking examples from nature inspire research efforts. This work presents novel theoretical and experimental drug discovery approaches for these challenges. In the first part, we introduce novel chemoinformatics approaches for clustering of large chemical libraries. The development of a fast algorithm for pairwise similarity calculations forms the basis for an exact and deterministic clustering method, which is able to process the available chemical space in a short time. We complement our chemoinformatics work by a novel approach for fast classification of small molecules according to the similarity of their frameworks, the so-called scaffolds. The method generates families of molecules that share geometry conserving scaffolds and we show that family members possess similar activity on identical targets. The second part introduces computational methods for PPI modulation. First, we present structure-based analysis of known stabilized PPIs, which enables the development of novel in silico approaches to screen for small molecule PPI stabilizers. We demonstrate their applicability by an experimentally tested virtual screening for 14-3-3 protein interaction stabilizers. Finally, we present a virtual screening approach dedicated to identify small molecule inhibitors of 14-3-3 protein interactions. Predicted inhibitors are experimentally verified and characterized by in vitro assays and X-ray crystallography. Structure-activity relationship studies yielded PPI inhibitors in the low micromolar range, which are also active in cell-based experiments.