Discovery and functional evaluation of diverse novel human CB1 receptor ligands

Machine Learning and Computational Chemistry for the Endocannabinoid System

Computational methods in medicinal chemistry facilitate drug discovery and design. In particular, machine learning methodologies have recently gained increasing attention. This chapter provides a structured overview of the current state of computational chemistry and its applications for the interrogation of the endocannabinoid system (ECS), highlighting methods in structure-based drug design, virtual screening, ligand-based quantitative structure-activity relationship (QSAR) modeling, and de novo molecular design. We emphasize emerging methods in machine learning and anticipate a forecast of future opportunities of computational medicinal chemistry for the ECS.

Energy windows for computed compound conformers: covering artefacts or truly large reorganization energies?

The generation of 3D conformers of small molecules underpins most computational drug discovery. Thus, the conformer quality is critical and depends on their energetics. A key parameter is the empirical conformational energy window (ΔEw), since only conformers within ΔEw are retained. However, ΔEw values in use appear unrealistically large. We analyze the factors pertaining to the conformer energetics and ΔEw. We argue that more attention must be focused on the problem of collapsed low-energy conformers. That is due to artificial intramolecular stabilization and occurs even with continuum solvation. Consequently, the conformational energy of extended bioactive structures is artefactually increased, which inflates ΔEw. Thus, this Perspective highlights the issues arising from low-energy conformers and suggests improvements via empirical or physics-based strategies.

Structure-Based Identification of Potent Natural Product Chemotypes as Cannabinoid Receptor 1 Inverse Agonists

Natural products are an abundant source of potential drugs, and their diversity makes them a rich and viable prospective source of bioactive cannabinoid ligands. Cannabinoid receptor 1 (CB1) antagonists are clinically established and well documented as potential therapeutics for treating obesity, obesity-related cardiometabolic disorders, pain, and drug/substance abuse, but their associated CNS-mediated adverse effects hinder the development of potential new drugs and no such drug is currently on the market. This limitation amplifies the need for new agents with reduced or no CNS-mediated side effects. We are interested in the discovery of new natural product chemotypes as CB1 antagonists, which may serve as good starting points for further optimization towards the development of CB1 therapeutics. In search of new chemotypes as CB1 antagonists, we screened the in silico purchasable natural products subset of the ZINC12 database against our reported CB1 receptor model using the structure-based virtual screening (SBVS) approach. A total of 18 out of 192 top-scoring virtual hits, selected based on structural diversity and key protein⁻ligand interactions, were purchased and subjected to in vitro screening in competitive radioligand binding assays. The in vitro screening yielded seven compounds exhibiting >50% displacement at 10 μM concentration, and further binding affinity (K_i and IC₅₀) and functional data revealed compound 16 as a potent and selective CB1 inverse agonist (K_i = 121 nM and EC₅₀ = 128 nM) while three other compounds-2, 12, and 18-were potent but nonselective CB1 ligands with low micromolar binding affinity (K_i). In order to explore the structure⁻activity relationship for compound 16, we further purchased compounds with >80% similarity to compound 16, screened them for CB1 and CB2 activities, and found two potent compounds with sub-micromolar activities. Most importantly, these bioactive compounds represent structurally new natural product chemotypes in the area of cannabinoid research and could be considered for further structural optimization as CB1 ligands.

Towards understanding the unbound state of drug compounds: Implications for the intramolecular reorganization energy upon binding

There has been an explosion of structural information for pharmaceutical compounds bound to biological targets, but the conformations and dynamics of compounds free in solution are poorly characterized, if at all. Yet, knowledge of the unbound state is essential to understand the fundamentals of molecular recognition, including the much debated conformational intramolecular reorganization energy of a compound upon binding (ΔEReorg). Also, dependable observation of the unbound compounds is important for ligand-based drug discovery, e.g. with pharmacophore modelling. Here, these questions are addressed with long (⩾0.5μs) state-of-the-art molecular dynamics (MD) simulations of 26 compounds (including 7 approved drugs) unbound in explicit solvent. These compounds were selected to be chemically diverse, with a range of flexibility, and good quality bioactive X-ray structures. The MD-simulated free compounds are compared to their bioactive structure and conformers generated with ad hoc sampling in vacuo or with implicit generalized Born (GB) aqueous solvation models. The GB conformational models clearly depart from those obtained in explicit solvent, and suffer from conformational collapse almost as severe as in vacuo. Thus, the global energy minima in vacuo or with GB are not suitable representations of the unbound state, which can instead be extensively sampled by MD simulations. Many, but not all, MD-simulated compounds displayed some structural similarity to their bioactive structure, supporting the notion of conformational pre-organization for binding. The ligand-protein complexes were also simulated in explicit solvent, to estimate ΔEReorg as an enthalpic difference ΔHReorg between the intramolecular energies in the bound and unbound states. This fresh approach yielded ΔHReorg values⩽6kcal/mol for 18 out of 26 compounds. For three particularly polar compounds 15⩽ΔHReorg⩽20kcal/mol, supporting the notion that ΔHReorg can be substantial. Those large ΔHReorg values correspond to a redistribution of electrostatic interactions upon binding. Overall, the study illustrates how MD simulations offer a promising avenue to characterize the unbound state of medicinal compounds.

Identifying the structural features and diversifying the chemical domain of peripherally acting CB1 receptor antagonists using molecular modeling techniques

A four featured pharmacophore and predictive 3D-QSAR models were developed which were used for virtual screening of the Asinex database to get chemically diverse hits of peripherally active CB1 receptor antagonists.

Current progress in Structure-Based Rational Drug Design marks a new mindset in drug discovery

The past decade has witnessed a paradigm shift in preclinical drug discovery with structure-based drug design (SBDD) making a comeback while high-throughput screening (HTS) methods have continued to generate disappointing results. There is a deficit of information between identified hits and the many criteria that must be fulfilled in parallel to convert them into preclinical candidates that have a real chance to become a drug. This gap can be bridged by investigating the interactions between the ligands and their receptors. Accurate calculations of the free energy of binding are still elusive; however progresses were made with respect to how one may deal with the versatile role of water. A corpus of knowledge combining X-ray structures, bioinformatics and molecular modeling techniques now allows drug designers to routinely produce receptor homology models of increasing quality. These models serve as a basis to establish and validate efficient rationales used to tailor and/or screen virtual libraries with enhanced chances of obtaining hits. Many case reports of successful SBDD show how synergy can be gained from the combined use of several techniques. The role of SBDD with respect to two different classes of widely investigated pharmaceutical targets: (a) protein kinases (PK) and (b) G-protein coupled receptors (GPCR) is discussed. Throughout these examples prototypical situations covering the current possibilities and limitations of SBDD are presented.

In silico investigation of interactions between human cannabinoid receptor-1 and its antagonists

Cannabinoid receptor-1 (CB(1)) is widely expressed in the central nervous system and plays a vital role in regulating food intake and energy expenditure. CB(1) antagonists such as Rimonabant have been used in clinic to inhibit food intake, and therefore reduce body weight in obese animals and humans. To investigate the binding modes of CB(1) antagonists to the receptor, both receptor- and ligand-based methods were implemented in this study. At first, a pharmacophore model was generated based on 31 diverse CB(1) antagonists collected from literature. A test set validation and a simulated virtual screening evaluation were then performed to verify the reliability and discriminating ability of the pharmacophore. Meanwhile, the homology model of CB(1) receptor was constructed based on the crystal structure of human β (2) adrenergic receptor (β (2)-AR). Several classical antagonists were then docked into the optimized homology model with induced fit docking method. A hydrogen bond between the antagonists and Lys192 on the third transmembrane helix of the receptor was formed in the docking study, which has proven to be critical for receptor-ligand interaction by biological experiments. The structure obtained from induced fit docking was then confirmed to be a reliable model for molecular docking from the result of the simulated virtual screening. The consistency between the pharmacophore and the homology structure further proved the previous observation. The built receptor structure and antagonists' pharmacophore should be useful for the understanding of inhibitory mechanism and development of novel CB(1) antagonists.

Rationalizing the role of SAR tolerance for ligand-based virtual screening

It is well appreciated that the results of ligand-based virtual screening (LBVS) are much influenced by methodological details, given the generally strong compound class dependence of LBVS methods. It is less well understood to what extent structure-activity relationship (SAR) characteristics might influence the outcome of LBVS. We have assessed the hypothesis that the success of prospective LBVS depends on the SAR tolerance of screening targets, in addition to methodological aspects. In this context, SAR tolerance is rationalized as the ability of a target protein to specifically interact with series of structurally diverse active compounds. In compound data sets, SAR tolerance articulates itself as SAR continuity, i.e., the presence of structurally diverse compounds having similar potency. In order to analyze the role of SAR tolerance for LBVS, activity landscape representations of compounds active against 16 different target proteins were generated for which successful LBVS applications were reported. In all instances, the activity landscapes of known active compounds contained multiple regions of local SAR continuity. When analyzing the location of newly identified LBVS hits and their SAR environments, we found that these hits almost exclusively mapped to regions of distinct local SAR continuity. Taken together, these findings indicate the presence of a close link between SAR tolerance at the target level, SAR continuity at the ligand level, and the probability of LBVS success.

State-of-the-art in ligand-based virtual screening

Virtual screening is a much discussed topic in chemoinformatics and medicinal chemistry, and widely applied in pharmaceutical research. Here, we provide an in-depth analysis of currently available ligand-based virtual screening applications. We formulate several scientific quality criteria for prospective ligand-based virtual screens and analyze, in detail, the information provided by currently available peer-reviewed publications. The results presented herein provide a detailed view of the current state-of-the-art in this field and point at several problematic issues but also opportunities for further advances.

Modeling approaches for ligand-based 3D similarity

3D ligand-based similarity approaches are widely used in the early phases of drug discovery for tasks such as hit finding by virtual screening or compound design with quantitative structure–activity relationships. Here in we review widely used software for performing such tasks. Some techniques are based on relatively mature technology, shape-based similarity for instance. Typically, these methods remained in the realm of the expert user, the experienced modeler. However, advances in implementation and speed have improved usability and allow these methods to be applied to databases comprising millions of compounds. There are now many reports of such methods impacting drug-discovery projects. As such, the medicinal chemistry community has become the intended market for some of these new tools, yet they may consider the wide array and choice of approaches somewhat disconcerting. Each method has subtle differences and is better suited to certain tasks than others. In this article we review some of the widely used computational methods via application, provide straightforward background on the underlying theory and provide examples for the interested reader to pursue in more detail. In the new era of preclinical drug discovery there will be ever more pressure to move faster and more efficiently, and computational approaches based on 3D ligand similarity will play an increasing role in in this process.

Probing the cannabinoid CB1/CB2 receptor subtype selectivity limits of 1,2-diarylimidazole-4-carboxamides by fine-tuning their 5-substitution pattern

The cannabinoid CB(1)/CB(2) receptor subtype selectivity in the 1,2-diarylimidazole-4-carboxamide series was boosted by fine-tuning its 5-substitution pattern. The presence of the 5-methylsulfonyl group in 11 led to a greater than approximately 840-fold CB(1)/CB(2) subtype selectivity. The compounds 10, 18 and 19 were found more active than rimonabant (1) in a CB(1)-mediated rodent hypotension model after oral administration. Our findings suggest a limited brain exposure of the P-glycoprotein substrates 11, 12 and 21.