Fragment-based drug discovery-the importance of high-quality molecule libraries

Abstract

Fragment-based drug discovery (FBDD) is now established as a complementary approach to high-throughput screening (HTS). Contrary to HTS, where large libraries of drug-like molecules are screened, FBDD screens involve smaller and less complex molecules which, despite a low affinity to protein targets, display more 'atom-efficient' binding interactions than larger molecules. Fragment hits can, therefore, serve as a more efficient start point for subsequent optimisation, particularly for hard-to-drug targets. Since the number of possible molecules increases exponentially with molecular size, small fragment libraries allow for a proportionately greater coverage of their respective 'chemical space' compared with larger HTS libraries comprising larger molecules. However, good library design is essential to ensure optimal chemical and pharmacophore diversity, molecular complexity, and physicochemical characteristics. In this review, we describe our views on fragment library design, and on what constitutes a good fragment from a medicinal and computational chemistry perspective. We highlight emerging chemical and computational technologies in FBDD and discuss strategies for optimising fragment hits. The impact of novel FBDD approaches is already being felt, with the recent approval of the covalent KRAS^G12C inhibitor sotorasib highlighting the utility of FBDD against targets that were long considered undruggable.

Keywords: covalent fragments; fraglites; fragment library; fragment-based drug discovery; machine learning; virtual screening.

Fig. 1

Timeline highlighting key papers influencing the course of FBDD (blue) and important dates showing its success (orange). In an early conceptual paper, Jencks outlined the additivity of binding energies for fragments of larger molecules [16]. Andrews et al. [17] subsequently estimated intrinsic binding energy contributions to ligand–receptor interactions for a range of functional groups. Based on a simple model of complementary ligand–receptor features, Hann et al. [18] proposed that molecules of lower complexity are likely to provide better starting points for drug discovery and discussed the need for highly sensitive assays. With increasing interest in fragment‐based drug discovery, commonly used metrics including ‘rule of three’ and ligand efficiency were developed [11, 19]. FBDD, fragment‐based drug discovery; FDA, United States Food and Drug Administration. [Colour figure can be viewed at wileyonlinelibrary.com]

Fig. 2

Property distributions in selected unfiltered large commercial fragment sets. (A) General fragment sets were obtained from Maybridge (30 061 compounds), Life Chemicals (50 597 compounds), Enamine (172 689 compounds) and ChemDiv (11 269 compounds). Hydrogen bond donors/acceptors, heavy atom count and total polar surface area including N, O, P and S were calculated in vortex software (Dotmatics, Bishops Stortford, UK). Predicted logD and water solubility were calculated in admet predictor software (Simulations Plus, Lancaster, CA, USA). Black lines denote mean or median for continuous or discrete properties, respectively. (B) Distributions of maximum internal similarity in the same fragment sets. For each compound, pairwise Tanimoto similarity was calculated in RDKit (RDKit: Open‐source cheminformatics; http://www.rdkit.org) against all other compounds in the set using Morgan fingerprints, radius 2. See the Section 2.2.2 for an explanation of Tanimoto similarity. For each compound, the maximum value of similarity against any other compound was retained. (C) Number of identical compounds in the same libraries. For example, 633 compounds are present in both Maybridge and Life Chemicals collections. HBA, hydrogen bond acceptor count; HBD, hydrogen bond donor count; HAC, heavy atom count; TPSA_NOPS, total polar surface area including N, O, P and S atoms; cLogD, calculated logarithm of distribution coefficient; LogSw, (calculated) logarithm of water solubility; maxSim, maximum internal similarity (as defined above). [Colour figure can be viewed at wileyonlinelibrary.com]

Fig. 3

Comparison of Cancer Research UK Beatson Drug Discovery Unit 1H fragment set against selected commercial sets. (A) Average maximum internal similarity. For each compound in any set, Tanimoto similarity was calculated against all other compounds in the set using Morgan fingerprints (radius 2) in RDKit. The maximum value was retained for each compound and averaged over the set. (B) Fraction of unique Bemis–Murcko scaffolds [29] in each set. Scaffold SMILES were extracted for each compound in each set using vortex software (Dotmatics, Bishops Stortford, UK) and unique canonical SMILES retained. The number of unique scaffolds was expressed relative to number of compounds in each library. Commercial sets: Enamine. High Fidelity Library (1920 compounds); Life Chemicals, General Fragment Library (50 607 compounds); ChemDiv, Fragments Library (11 269 compounds); Bionet, 2nd Generation Premium Library (1166 compounds). SMILES, simplified molecular‐input line‐entry system. [Colour figure can be viewed at wileyonlinelibrary.com]