Bioactivity-Guided Navigation of Chemical Space

摘要

Acentral aim of biological research is to elucidate the many roles of proteins in complex, dynamic living systems; thenselective perturbation of protein function is an important tool in achieving this goal. Because chemical perturbationsnoffer opportunities often not accessible with genetic methods, the development of small-molecule modulators of protein func-ntion is at the heart of chemical biology research. In this endeavor, the identification of biologically relevant starting pointsnwithin the vast chemical space available for the design of compound collections is a particularly relevant, yet difficult, task.nIn this Account, we present our research aimed at linking chemical and biological space to define suitable starting pointsnthat guide the synthesis of compound collections with biological relevance.nBoth protein folds and natural product (NP) scaffolds are highly conserved in nature. Whereas different amino acid sequencesncan make up ligand-binding sites in proteins with highly similar fold types, differently substituted NPs characterized by particularnscaffold classes often display diverse biological activities. Therefore, we hypothesized that (i) ligand-binding sites with similar ligand-nsensing cores embedded in their folds would bind NPs with similar scaffolds and (ii) selectivity is ensured by variation of bothnamino acid side chains and NP substituents. To investigate this notion in compound library design, we developed an approachntermed biology-oriented synthesis (BIOS). BIOS employs chem- and bioinformatic methods for mapping biologically relevant chem-nical space and protein space to generate hypotheses for compound collection design and synthesis. BIOS also provides hypoth-neses for potential bioactivity of compound library members. On the one hand, protein structure similarity clustering (PSSC) is usednto identify ligand binding sites with high subfold similarity, that is, high structural similarity in their ligand-sensing cores. On thenother hand, structural classification by scaffold trees (for example, structural classification of natural products or SCONP), whenncombined with software tools like “Scaffold Hunter”, enables the hierarchical structural classification of small-molecule collectionsnin tree-like arrangements, their annotation with bioactivity data, and the intuitive navigation of chemical space. Brachiation (in anmanner analogous to tree-swinging primates) within the scaffold trees serves to identify new starting points for the design andnsynthesis of small-molecule libraries, and PSSC may be used to select potential protein targets.