Fragment-Based Drug Discovery by NMR. Where Are the Successes and Where can It Be Improved?

Abstract

Over the last century, the definitions of pharmaceutical drug and drug discovery have changed considerably. Evolving from an almost exclusively serendipitous approach, drug discovery nowadays involves several distinct, yet sometimes interconnected stages aimed at obtaining molecules able to interact with a defined biomolecular target, and triggering a suitable biological response. At each of the stages, a wide range of techniques are typically employed to obtain the results required to move the project into the next stage. High Throughput Screening (HTS) and Fragment Based Drug Design (FBDD) are the two main approaches used to identify drug-like candidates in the early stages of drug discovery. Nuclear Magnetic Resonance (NMR) spectroscopy has many applications in FBDD and is used extensively in industry as well as in academia. In this manuscript, we discuss the paths of both successful and unsuccessful molecules where NMR had a crucial part in their development. We specifically focus on the techniques used and describe strengths and weaknesses of each stage by examining several case studies. More precisely, we examine the development history from the primary screening to the final lead optimisation of AZD3839 interacting with BACE-1, ABT-199 interacting with BCL_2/XL and S64315 interacting with MCL-1. Based on these studies, we derive observations and conclusions regarding the FBDD process by NMR and discuss its potential improvements.

Keywords: BACE-1; BCL-2; Fragments Based Drug Discovery; MCL-1; NMR-FBDD; Venetoclax.

FIGURE 1

Usage of FBDD methods in the development of new molecules. (A) Total count of journal articles from Jan-2015 to Dec-2020 retrieved by querying “NMR and Fragment-based Drug discovery” in the PubMed Central database (see Supplementary Materials for the conditional query script). (B) Total count of FDA-approved New Molecular Entities (NMEs) and original biologics for the same time range. In dark blue, the small molecule NMEs, including the three known drugs whose fragment origins were derived from NMR studies. In light blue, the approved biologics, such as antibodies and oligonucleotides, and other chemicals entities such as diagnostics, combinations of old drugs, natural products. (C) Normalised scores of the occurrences of NMR spectroscopy as a technique in the discovery and development of molecules across the inspected cases. (D) Normalised scores for the total count of the various NMR techniques used throughout the drug discovery process of the inspected cases. Note that for both (C,D), some compounds have been through multiple stages of drug discovery by NMR, so fractions sum to values >1.0.

FIGURE 2

The AZD-3839 case-study. (A) The optimisation pathway: from NMR hits to AZD-3839. Compound-1 represents the hit initially identified from the Water-LOGSY NMR study. The blue circle highlights the crucial isocytosine aromatic proton. Compounds were optimised through a series of crystallography-based methods to yield the final compound-8 (AZD-3839) (Geschwindner et al., 2007; Jeppsson et al., 2012), yet preserving the original amidine motif (red circle) already present in compound-1. (B) Molecular structure representation of BACE-1 (PDB code: 4B05) and the main interaction between the catalytic groove (Asp32 and Asp228) and the amidine group of AZD-3839, first observed in the NMR-discovered hit (black rectangle). (C) Molecular similarity, as expressed by the Tanimoto coefficient (MS, Blue), normalised molecular weight (MW, orange) and polar surface area (PSA, green) scores for the eight compounds on the development path of AZD-3839.

FIGURE 3

The ABT-199 case-study. (A) Optimisation pathway: from NMR hits to ABT-199. The aromatic moieties of the NMR-determined hits (green and cyan circles) were originally identified as interacting with BCL-XL active sites (Petros et al., 2006). Compounds 1, 2, 3, 4 were identified and optimised through NMR methodologies, whereas the latest ABT compounds optimisations benefited from X-ray crystallography techniques (Petros et al., 2006; Petros et al., 2010; Souers et al., 2013). (B) Molecular structure representation of BCL-2 in complex with Venetoclax (PDB code: 6O0L). Green and cyan circles indicated the aromatic motifs originally identified through the NMR primary screening. (C) Molecular similarity, as expressed by the Tanimoto coefficient (MS, Blue), normalised molecular weight (MW, orange) and polar surface area (PSA, green) scores for compounds 3-7 on the development path of ABT-199.

FIGURE 4

The S64315 case-study. (A) Optimisation pathway: from NMR hits to S64315. All compound nomenclatures are identical to those used in the original manuscript (Szlávik et al., 2019) for an easier comparability. Compound 1a represents the initially identified thienopyrimidine core by ligand-detected 1D NMR techniques. The green and blue circles for compound 5d highlight the chemical groups that gave rise to crucial NOEs that suggested the initial molecule binding poses (Szlávik et al., 2019). (B) Molecular structure representation of a model of MCL-1 in complex with compound 18a (PDB code: 6QYO). The green ellipse highlights the original thienopyrimidine motif first identified by an 1D-NMR screening experiment (Szlávik et al., 2019). (C) Molecular similarity, as expressed by the Tanimoto coefficient (MS, Blue), normalised molecular weight (MW, orange) and polar surface area (PSA, green) scores for the twelve compounds on the development path of S64315 AZD-3839.

FIGURE 5

Comparison of methods and history of clinical drugs. (A) Normalised score of the predominant methodologies used for the discovery and development of the 53 clinical drugs inspected in this study. (B) Normalised similarity scores for the ABT-199 (orange), the S64315 (green) and the AZD-3839 (blue) fragment-to-drug developments pathways with interpolated optimisations steps.