Macrocycles as inhibitors of protein-protein interactions
Embargo Date
2023-11-01
OA Version
Citation
Abstract
Among the many drugs that have been developed from natural products, macrocyclic compounds (MCs) – compounds containing a ring of 11 or more atoms – are a recurring chemotype. It is not fully understood how large MCs can have utility as drugs, despite violating conventional notions of druglikness. In recent years, there has been great interest in the potential utility of MCs as drugs, especially for modulation of traditionally challenging targets, such as protein-protein interactions, where high molecular weight compounds or other non-canonical chemotypes might be required for high affinity binding. As a result of this interest, there has been considerable recent effort towards understanding which specific molecular features of MCs might account for their advantageous properties. In this thesis, I take several approaches to better understand macrocycles as a chemotype, and their application for inhibiting challenging protein-protein interaction drug targets.
Using the protein-protein interaction between β-catenin and T-cell factor 4 (Tcf4) as a test system, I explore whether macrocycles are advantageous for inhibiting challenging protein-protein interaction targets. I describe the optimization of fluorescence anisotropy binding and inhibition assays for this protein pair, and how the assay was adapted for high throughput screening. The results of screening both small molecule compounds and MCs for inhibition of the β-catenin/Tcf4 interaction are included. The hits identified from these screens were validated in dose-response experiments, and the small molecule hit compounds were computationally docked to β-catenin. In parallel, I address the feasibility of designing MCs for inhibition of the β-catenin/Tcf4 interaction using rationally designed cyclic peptides. The results obtained using linear peptides as preliminary probes are described, as well as the design of and synthetic strategies for cyclic peptide inhibitors.
To better understand what features of macrocycles are important to confer good pharmaceutical properties, I take cheminformatic and computational approaches to compare oral and non-oral MC drugs and clinical candidates with commercially available synthetic MC collections. A major part of this work was to develop novel macrocycle-specific structural descriptors, and validate that they are useful for explaining the unique features of macrocycle compounds. I performed a principal component analysis and I identify three compact, yet distinct, regions of structure-property space where oral MC drugs and clinical candidates reside. A set of guidelines was developed to provide direction for how to design novel synthetic MCs that possess physicochemical properties resembling those of oral macrocycle drugs. Lastly, I apply the machine learning method of partial least squares regression to build predictive models for passive membrane permeability of MCs, using the novel macrocycle-specific property descriptors I developed. Methods to build these models were tested and optimized, and preliminary predictive models are presented. Preliminary results highlight which properties of MCs could be incorporated into synthetic compound designs to confer improved prospects for passive membrane permeability.