Navigating 3D Chemical Space
Abstract
In the late 1990s, fragment-based drug discovery (FBDD) emerged as a novel method for hit identification, complementing existing strategies such as high-throughput screening (HTS). Rather than screening drug-sized molecules, FBDD focuses on smaller fragment-sized compounds with molecular weights of up to 300 Da. Owing to their reduced complexity, fragments sample chemical space efficiently, which allows for the use of significantly smaller libraries that typically comprise only a few thousand compounds. Historically, fragment libraries have been dominated by relatively flat, (hetero)aromatic scaffolds. Although such libraries have successfully yielded several clinically approved drugs and numerous clinical candidates, 2D fragments explore only a narrow region of chemical space and may bias discovery toward planar chemotypes with suboptimal three-dimensional fit. In contrast, 3D fragments offer greater diversity in terms of shape and exit vectors, which, depending on the target, may better complement the three-dimensional nature of protein binding sites. This has led to growing interest in incorporating 3D fragments alongside traditional 2D libraries. However, 3D fragments also present challenges, including reduced synthetic accessibility and potentially lower hit rates. This thesis explores the challenges, as well as the opportunities, associated with 3D fragments across the key stages of fragment-based drug discovery. Following an introduction in Chapter 1, the thesis commences in Chapter 2 with a comparative analysis of the metrics commonly used to quantify the three-dimensional character of fragments, including the fraction of sp3-hybridized carbons (Fsp3), Plane of Best Fit (PBF), and Principal Moments of Inertia (PMI). Implications for their use in (fragment) library design are discussed. Chapter 3 describes the development of an automated, open-source workflow for fragment library design. The workflow is highly customizable and accounts for chemical diversity and novelty, but also allows for the incorporation of the 3D metrics evaluated in Chapter 2. To demonstrate its utility, we used it to design and synthesize a focused set of cyclopropane-based 3D fragments as isolated cis- and trans-isomers. Chapter 4 details the synthesis of two focused libraries containing a total of 28 spirocyclic cyclobutane fragments, designed using the automated workflow described in Chapter 3. It describes the large-scale synthesis of diastereomerically pure building blocks and their parallel chemistry-based derivatization into the two focused libraries, each composed of isolated sets of achiral diastereomers. Chapter 5 describes the identification of a cyclobutane-based 3D fragment hit targeting the Histamine H1 receptor (H1R), discovered by screening a diverse 80-membered set of 3D fragments comprising various aliphatic (hetero)cyclic scaffolds. Following several rounds of SAR-driven optimization, the high-affinity antagonist VUF26691 (pKi = 8.8) was identified, with retention of the cyclobutane core throughout the optimization. Chapter 6 describes the development of the Exit Vector Fingerprint (EFVP) as a novel method for designing 3D fragment libraries. Rather than relying on overall molecular shape or atom hybridization, it employs exit vector geometry, combined with the pharmacophoric features residing on those vectors, as a more pharmacologically relevant basis for describing three-dimensionality. In addition to providing a practical tool for designing the next generation of 3D fragment libraries, the EVFP offers a refined conceptual framework for evaluating three-dimensional fragments. Finally, Chapter 7 provides an integrated discussion of the findings presented throughout the thesis and outlines future perspectives.