Alternative quality assessment strategy to compare performances of GPCR-ligand docking protocols: the human adenosine A(2A) receptor as a case study

PoseBusters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences

The last few years have seen the development of numerous deep learning-based protein-ligand docking methods. They offer huge promise in terms of speed and accuracy. However, despite claims of state-of-the-art performance in terms of crystallographic root-mean-square deviation (RMSD), upon closer inspection, it has become apparent that they often produce physically implausible molecular structures. It is therefore not sufficient to evaluate these methods solely by RMSD to a native binding mode. It is vital, particularly for deep learning-based methods, that they are also evaluated on steric and energetic criteria. We present PoseBusters, a Python package that performs a series of standard quality checks using the well-established cheminformatics toolkit RDKit. The PoseBusters test suite validates chemical and geometric consistency of a ligand including its stereochemistry, and the physical plausibility of intra- and intermolecular measurements such as the planarity of aromatic rings, standard bond lengths, and protein-ligand clashes. Only methods that both pass these checks and predict native-like binding modes should be classed as having "state-of-the-art" performance. We use PoseBusters to compare five deep learning-based docking methods (DeepDock, DiffDock, EquiBind, TankBind, and Uni-Mol) and two well-established standard docking methods (AutoDock Vina and CCDC Gold) with and without an additional post-prediction energy minimisation step using a molecular mechanics force field. We show that both in terms of physical plausibility and the ability to generalise to examples that are distinct from the training data, no deep learning-based method yet outperforms classical docking tools. In addition, we find that molecular mechanics force fields contain docking-relevant physics missing from deep-learning methods. PoseBusters allows practitioners to assess docking and molecular generation methods and may inspire new inductive biases still required to improve deep learning-based methods, which will help drive the development of more accurate and more realistic predictions.

Therapeutic properties and molecular docking study of some phenolic compounds as anti-human lung cancer potential: A biochemical approach

Chloroxine (5,7-dichloro-8-hydroxyquinoline) is a molecule utilized in some shampoos for the therapy of seborrheic dermatitis of the scalp and dandruff. In this study, we investigated the inhibition effects of 5,7-dichloro-8-hydroxyquinoline and methyl 3,4,5-trihydroxybenzoate compounds on the 3-hydroxy-3-methylglutaryl coenzyme-A (HMG-CoA Reductase) and urease enzymes. We have obtained results for the HMG-CoA Reductase and urease enzymes at the micromolar level. In our study, inhibition result of 5,7-dichloro-8-hydroxyquinoline and Methyl 3,4,5-trihydroxybenzoate on HMG-CoA reductase showed lower values 2.28 ± 0.78 and 33.25 ± 5.04 µg/ml, respectively. Additionally, inhibition result of 5,7-dichloro-8-hydroxyquinoline and methyl 3,4,5-trihydroxybenzoate on urease showed lower values 6.18 ± 1.38 and 8.51 ± 1.35 µg/ml, respectively. Molecular docking calculations were made for their biological activities were compared. In the present work, the structures of the related compounds (1 and 2) were drawn using Gaussian 09 software and done geometry optimization at DFT/B3LYP/6-31G* basis set with aforementioned program. Cytotoxicity potential of these compounds against human lung cancer demonstrated that these compounds had good cytotoxic effects. Both compounds significantly decreased lung cell viability from low doses. In addition, 100 µM dose of all compounds caused significant reductions in lung cell viability. In general, we can say that of the two tested compounds, 5,7-dichloro-8-hydroxyquinoline and methyl 3,4,5-trihydroxybenzoate have cytotoxic effects in all cell types, and this effect is particularly strong in lung cells. Activities were performed at concentrations of 10, 20, 50, 70, and 100 µl and we achieved good results. Lung cell viability (%) value was better at 100 µl concentration and IC₅₀ of them were 54.28 and 48.05 µM.

One class classification for the detection of β2 adrenergic receptor agonists using single-ligand dynamic interaction data

G protein-coupled receptors are involved in many biological processes, relaying the extracellular signal inside the cell. Signaling is regulated by the interactions between receptors and their ligands, it can be stimulated by agonists, or inhibited by antagonists or inverse agonists. The development of a new drug targeting a member of this family requires to take into account the pharmacological profile of the designed ligands in order to elicit the desired response. The structure-based virtual screening of chemical libraries may prioritize a specific class of ligands by combining docking results and ligand binding information provided by crystallographic structures. The performance of the method depends on the relevance of the structural data, in particular the conformation of the targeted site, the binding mode of the reference ligand, and the approach used to compare the interactions formed by the docked ligand with those formed by the reference ligand in the crystallographic structure. Here, we propose a new method based on the conformational dynamics of a single protein–ligand reference complex to improve the biased selection of ligands with specific pharmacological properties in a structure-based virtual screening exercise. Interactions patterns between a reference agonist and the receptor, here exemplified on the β2 adrenergic receptor, were extracted from molecular dynamics simulations of the agonist/receptor complex and encoded in graphs used to train a one-class machine learning classifier. Different conditions were tested: low to high affinity agonists, varying simulation duration, considering or ignoring hydrophobic contacts, and tuning of the classifier parametrization. The best models applied to post-process raw data from retrospective virtual screening obtained by docking of test libraries effectively filtered out irrelevant poses, discarding inactive and non-agonist ligands while identifying agonists. Taken together, our results suggest that consistency of the binding mode during the simulation is a key to the success of the method.

Pellucidin A promotes antinociceptive activity by peripheral mechanisms inhibiting COX-2 and NOS: In vivo and in silico study

Peperomia pellucida (PP) belongs to the Peperomia genus, which has a pantropic distribution. PP is used to treat a wide range of symptoms and diseases, such as pain, inflammation, and hypertension. Intriguingly, PP extract is used by different tropical countries for its anti-inflammatory and antinociceptive effects. In fact, these outcomes have been shown in animal models, though the exact bioactive products of PP that exert such results are yet to be discovered. To determine and elucidate the mechanism of action of one of these compounds, we evaluated the antinociceptive effect of the novel dimeric ArC2 compound, Pellucidin A by using in vivo and in silico models. Animals were then subjected to chemical, biphasic and thermal models of pain. Pellucidin A induced an antinociceptive effect against chemical-induced pain in mice, demonstrated by the decrease of the number of writhes, reaching a reduction of 43% and 65% in animals treated with 1 and 5 mg/kg of Pellucidin A, respectively. In the biphasic response (central and peripheral), animals treated with Pellucidin A showed a significant reduction of the licking time exclusively during the second phase (inflammatory phase). In the hot-plate test, Pellucidin A did not have any impact on the latency time of the treated animals. Moreover, in vivo and in silico results show that Pellucidin A’s mechanism of action in the inflammatory pain occurs most likely through interaction with the nitric oxide (NO) pathway. Our results demonstrate that the antinociceptive activities of Pellucidin A operate under mechanism(s) of peripheral action, involving inflammatory mediators. This work provides insightful novel evidence of the biological properties of Pellucidin A, and leads to a better understanding of its mechanism of action, pointing to potential pharmacological use.

Ligand Pose Predictions for Human G Protein-Coupled Receptors: Insights from the Amber-Based Hybrid Molecular Mechanics/Coarse-Grained Approach

Human G protein-coupled receptors (hGPCRs) are the most frequent targets of Food and Drug Administration (FDA)-approved drugs. Structural bioinformatics, along with molecular simulation, can support structure-based drug design targeting hGPCRs. In this context, several years ago, we developed a hybrid molecular mechanics (MM)/coarse-grained (CG) approach to predict ligand poses in low-resolution hGPCR models. The approach was based on the GROMOS96 43A1 and PRODRG united-atom force fields for the MM part. Here, we present a new MM/CG implementation using, instead, the Amber 14SB and GAFF all-atom potentials for proteins and ligands, respectively. The new implementation outperforms the previous one, as shown by a variety of applications on models of hGPCR/ligand complexes at different resolutions, and it is also more user-friendly. Thus, it emerges as a useful tool to predict poses in low-resolution models and provides insights into ligand binding similarly to all-atom molecular dynamics, albeit at a lower computational cost.

A Deep-Learning Approach toward Rational Molecular Docking Protocol Selection

While a plethora of different protein–ligand docking protocols have been developed over the past twenty years, their performances greatly depend on the provided input protein–ligand pair. In this study, we developed a machine-learning model that uses a combination of convolutional and fully connected neural networks for the task of predicting the performance of several popular docking protocols given a protein structure and a small compound. We also rigorously evaluated the performance of our model using a widely available database of protein–ligand complexes and different types of data splits. We further open-source all code related to this study so that potential users can make informed selections on which protocol is best suited for their particular protein–ligand pair.

Can We Still Trust Docking Results? An Extension of the Applicability of DockBench on PDBbind Database

The number of entries in the Protein Data Bank (PDB) has doubled in the last decade, and it has increased tenfold in the last twenty years. The availability of an ever-growing number of structures is having a huge impact on the Structure-Based Drug Discovery (SBDD), allowing investigation of new targets and giving the possibility to have multiple structures of the same macromolecule in a complex with different ligands. Such a large resource often implies the choice of the most suitable complex for molecular docking calculation, and this task is complicated by the plethora of possible posing and scoring function algorithms available, which may influence the quality of the outcomes. Here, we report a large benchmark performed on the PDBbind database containing more than four thousand entries and seventeen popular docking protocols. We found that, even in protein families wherein docking protocols generally showed acceptable results, certain ligand-protein complexes are poorly reproduced in the self-docking procedure. Such a trend in certain protein families is more pronounced, and this underlines the importance in identification of a suitable protein–ligand conformation coupled to a well-performing docking protocol.

Could the presence of sodium ion influence the accuracy and precision of the ligand-posing in the human A2A adenosine receptor orthosteric binding site using a molecular docking approach? Insights from Dockbench

The allosteric modulation of G protein-coupled receptors (GPCRs) by sodium ions has received considerable attention as crystal structures of several receptors, in their inactive conformation, show a Na⁺ ion bound to specific residues which, in the human A_2A adenosine receptor (hA_2A AR), are Ser91^3.39, Trp246^6.48, Asn280^7.45, and Asn284^7.49. A cluster of water molecules completes the coordination of the sodium ion in the putative allosteric site. It is absolutely consolidated that the progress made in the field of GPCRs structural determination has increased the adoption of docking-driven approaches for the identification or the optimization of novel potent and selective ligands. Despite the extensive use of docking protocols in virtual screening approaches, to date, almost any of these studies have been carried out without taking into account the presence of the sodium cation and its first solvation shell in the putative allosteric binding site. In this study, we have focused our attention on determining how the presence of sodium ion binding and additionally its first hydration sphere, in hA_2AAR could influence the ligand positioning accuracy during molecular docking simulations for most of the available resting and activated hA_2A AR crystal structures, using DockBench as a comparative benchmarking tool and implementing a new correlation coefficient (EM). This work provides indications on the evidence that the posing performance (accuracy and/or precision) of the docking protocols in reproducing the crystallographic poses of different hA_2A AR antagonists is generally increased in the presence of the sodium cation and its first solvation shell, in agreement with experimental observations. Consequently, the inclusion of sodium ion and its first solvation shell should be considered in order to facilitate the selection of new potential ligands in all molecular docking-based virtual screening protocols that aim to find novel GPCRs antagonists and inverse agonists.

Discovery of 2-aminoimidazole and 2-amino imidazolyl-thiazoles as non-xanthine human adenosine A3 receptor antagonists: SAR and molecular modeling studies

A small-molecule combinatorial library of 24 compounds with 2-aminoimidazole and 2-aminoimidazolyl-thiazole derivatives was synthesized using a 2-chloro trityl resin. The generated compound library was tested against all the human adenosine receptors subtypes. The 2-aminoimidazole derivatives (6a-6l) showed weak to moderate affinity towards the human adenosine receptors. Further modification to 2-aminoimidazolyl-thiazole derivatives (12a-12l) resulted in an improvement of affinity at adenosine A₁, A_2A and A₃ receptor subtypes. Compound 12b was the most potent and selective non-xanthine human adenosine A₃ receptor antagonist of this series. A receptor-based modeling study was performed to explore the possible binding mode of these novel 2-aminoimidazole and 2-aminoimidazolyl-thiazole derivatives into human adenosine A₁, A_2A and A₃ receptor subtypes.

Comparison of the Human A2A Adenosine Receptor Recognition by Adenosine and Inosine: New Insight from Supervised Molecular Dynamics Simulations

Adenosine deaminase converts adenosine into inosine. In contrast to adenosine, relatively little attention has been paid to the physiological roles of inosine. Nevertheless, recent studies have demonstrated that inosine has neuroprotective, cardioprotective, immunomodulatory, and antidepressive effects. Inosine was recently shown to be a less potent agonist than adenosine at the A_2A adenosine receptor. To better depict the differences in the mechanisms of receptor recognition between adenosine and inosine, we carried out supervised molecular dynamics (SuMD) simulations, and the results are analyzed herein.

Structural features embedded in G protein-coupled receptor co-crystal structures are key to their success in virtual screening

Structure based drug discovery on GPCRs harness atomic detail X-ray binding pockets and large libraries of potential drug lead candidates in virtual screening (VS) to identify novel lead candidates. Relatively small conformational differences between such binding pockets can be critical to the success of VS. Retrospective VS on GPCR/ligand co-crystal structures revealed stark differences in the ability of different structures to identify known ligands, despite being co-crystallized with the same ligand. When using the OpenEye toolkit and the ICM modeling package, we identify criteria associated with the predictive power of binding pockets in VS that consists of a combination of ligand/receptor interaction pattern and predicted ligand/receptor interaction strength. These findings can guide the selection and refinement of GPCR binding pockets for use in SBDD programs and may also provide a potential framework for evaluating the ability of computational GPCR binding pocket refinement tools in improving the predictive power of binding pockets.

The role of 5-arylalkylamino- and 5-piperazino- moieties on the 7-aminopyrazolo[4,3-d]pyrimidine core in affecting adenosine A1 and A2A receptor affinity and selectivity profiles

New 7-amino-2-phenylpyrazolo[4,3-d]pyrimidine derivatives, substituted at the 5-position with aryl(alkyl)amino- and 4-substituted-piperazin-1-yl- moieties, were synthesized with the aim of targeting human (h) adenosine A₁ and/or A_2A receptor subtypes. On the whole, the novel derivatives 1–24 shared scarce or no affinities for the off-target hA_2B and hA₃ ARs. The 5-(4-hydroxyphenethylamino)- derivative 12 showed both good affinity (K_i = 150 nM) and the best selectivity for the hA_2A AR while the 5-benzylamino-substituted 5 displayed the best combined hA_2A (K_i = 123 nM) and A₁ AR affinity (K_i = 25 nM). The 5-phenethylamino moiety (compound 6) achieved nanomolar affinity (K_i = 11 nM) and good selectivity for the hA₁ AR. The 5-(N⁴-substituted-piperazin-1-yl) derivatives 15–24 bind the hA₁ AR subtype with affinities falling in the high nanomolar range. A structure-based molecular modeling study was conducted to rationalize the experimental binding data from a molecular point of view using both molecular docking studies and Interaction Energy Fingerprints (IEFs) analysis.

New Trends in Inspecting GPCR-ligand Recognition Process: the Contribution of the Molecular Modeling Section (MMS) at the University of Padova

In this review, we present a survey of the recent advances carried out by our research groups in the field of ligand-GPCRs recognition process simulations recently implemented at the Molecular Modeling Section (MMS) of the University of Padova. We briefly describe a platform of tools we have tuned to aid the identification of novel GPCRs binders and the better understanding of their binding mechanisms, based on two extensively used computational techniques such as molecular docking and MD simulations. The developed methodologies encompass: (i) the selection of suitable protocols for docking studies, (ii) the exploration of the dynamical evolution of ligand-protein interaction networks, (iii) the detailed investigation of the role of water molecules upon ligand binding, and (iv) a glance at the way the ligand might go through prior reaching the binding site.

5,7-Disubstituted-[1,2,4]triazolo[1,5-a][1,3,5]triazines as pharmacological tools to explore the antagonist selectivity profiles toward adenosine receptors

The structure-activity relationship of new 5,7-disubstituted-[1,2,4]triazolo[1,5-a][1,3,5]triazines as adenosine receptors (ARs) antagonists has been explored. The introduction of a benzylamino group at C5 with a free amino group at C7 increases the affinity toward all the ARs subtypes (10: KihA1 = 94.6 nM; KihA2A = 1.11 nM; IC50hA2B = 2214 nM; KihA3 = 30.8 nM). Replacing the free amino group at C7 with a phenylureido moiety yields a potent and quite selective hA2A AR antagonist (14: hA2A AR Ki = 1.44 nM; hA1/hA2A = 216.0; hA3/hA2A = 20.6). This trend diverges from the analysis on the pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine series previously reported. With the help of an in silico receptor-driven approach, we have rationalized these observations and elucidated from a molecular point of view the role of the benzylamino group at C5 in determining affinity toward the hA2A AR.

The Influence of the 1-(3-Trifluoromethyl-Benzyl)-1H-Pyrazole-4-yl Moiety on the Adenosine Receptors Affinity Profile of Pyrazolo[4,3-e][1,2,4]Triazolo[1,5-c]Pyrimidine Derivatives

A new series of pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine (PTP) derivatives has been developed in order to explore their affinity and selectivity profile at the four adenosine receptor subtypes. In particular, the PTP scaffold was conjugated at the C2 position with the 1-(3-trifluoromethyl-benzyl)-1H-pyrazole, a group believed to confer potency and selectivity toward the human (h) A_2B adenosine receptor (AR) to the xanthine ligand 8-(1-(3-(trifluoromethyl)benzyl)-1H-pyrazol-4-yl)-1,3-dimethyl-1H-purine-2,6(3H,7H)-dione (CVT 6975). Interestingly, the synthesized compounds turned out to be inactive at the hA_2B AR but they displayed affinity at the hA₃ AR in the nanomolar range. The best compound of the series (6) shows both high affinity (hA₃ AR K_i = 11 nM) and selectivity (A₁/A₃ and A_2A/A₃ > 9090; A_2B/A₃ > 909) at the hA₃ AR. To better rationalize these results, a molecular docking study on the four AR subtypes was performed for all the synthesized compounds. In addition, CTV 6975 and two close analogues have been subjected to the same molecular docking protocol to investigate the role of the 1-(3-trifluoromethyl-benzyl)-1H-pyrazole on the binding at the four ARs.

DockBench: An Integrated Informatic Platform Bridging the Gap between the Robust Validation of Docking Protocols and Virtual Screening Simulations

Virtual screening (VS) is a computational methodology that streamlines the drug discovery process by reducing costs and required resources through the in silico identification of potential drug candidates. Structure-based VS (SBVS) exploits knowledge about the three-dimensional (3D) structure of protein targets and uses the docking methodology as search engine for novel hits. The success of a SBVS campaign strongly depends upon the accuracy of the docking protocol used to select the candidates from large chemical libraries. The identification of suitable protocols is therefore a crucial step in the setup of SBVS experiments. Carrying out extensive benchmark studies, however, is usually a tangled task that requires users' proficiency in handling different file formats and philosophies at the basis of the plethora of existing software packages. We present here DockBench 1.0, a platform available free of charge that eases the pipeline by automating the entire procedure, from docking benchmark to VS setups. In its current implementation, DockBench 1.0 handles seven docking software packages and offers the possibility to test up to seventeen different protocols. The main features of our platform are presented here and the results of the benchmark study of human Checkpoint kinase 1 (hChk1) are discussed as validation test.