Computational tools for the discovery of novel aromatic heterocyclic bioisosteres

Thesis

Aromatic heterocycles are a structural feature of significant importance in small molecule therapeutics; indeed over 85% of the small molecules approved by the FDA between 2020–2024 contain at least one aromatic heterocycle. Key to their utility in medicinal chemistry is their rigid geometric projection of important functionality, the sensitivity with which the physicochemical properties of the parent molecule can be tuned by altering the constitution of the heterocycle, and the range of reliable methodologies enabling their synthesis. Consequently, this structural and physicochemical versatility renders them useful in bioisosteric replacement strategies, where a chemical moiety in a molecule is substituted for another to improve one or more drug-like properties, whilst preserving the biological profile of the initial compound.

Despite the widespread utility of aromatic heterocycles, the proportion of aromatic heterocyclic chemical space that is regularly sampled in medicinal chemistry is limited, and thus constrains the drug-like property space available to drug discovery campaigns. Virtual libraries of compounds enumerated such that they contain large regions of previously unsynthesised molecules have been developed, but tools to explore these for biologically-relevant applications are limited. The work described in this thesis employs the widely-accepted definitions of bioisosterism, relating similarities between molecular shape and electron distribution to broader biological effects, to design and evaluate computational tools that explore regions of aromatic heterocyclic chemical space for new bioisosteres of commonly-occurring heterocycles in medicinal chemistry.

Chapter 1 introduces the relevance of small molecules in drug discovery, and the reasons for the popularity of aromatic heterocycles within these important modalities in medicinal chemistry. An overview of previous work in the field of molecular similarity searching and bioisostere discovery is also presented. Chapter 2 describes the development and implementation of the first generation of the Heterocycle Isostere Explorer (HCIE), and its merits and limitations are discussed. In Chapter 3, the creation of the MoBiVic library of mono- and bifunctionalised aromatic heterocycles is described, leading to the development and implementation of the second generation of HCIE in Chapter 4, built around a unique, vector-based alignment algorithm, and a new implementation of electrostatic and shape similarity scoring. The last two chapters describe ongoing experimental work to validate the results of the tools described in the earlier chapters, with Chapter 5 exploring synthetic strategies for accessing novel aromatic heterocycles, and discussing the challenges encountered in their preparation. Chapter 6 describes the role of HCIE and other computational tools in ongoing medicinal chemistry projects at the Centre for Medicines Discovery.

Authors: Holland, MTO

• DOI: 10.5287/ora-dmd9bqaoj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: medicinal chemistry; drug discovery; software development; computational drug discovery
• Subjects: Chemistry; Pharmaceutical chemistry; Medical sciences