Alchemical approach performance in calculating the ligand-binding free energy.

Alchemical binding free energy calculations are one of the most accurate methods for estimating ligand-binding affinity. Assessing the accuracy of the approach over protein targets is one of the most interesting issues. The free energy difference of binding between a protein and a ligand was calculated via the alchemical approach. The alchemical approach exhibits satisfactory accuracy over four targets, including AmpC beta-lactamase (AmpC); glutamate receptor, ionotropic kainate 1 (GluK1); heat shock protein 90 (Hsp90); and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) main protease (Mpro). In particular, the correlation coefficients between calculated binding free energies and the respective experiments over four targets range from 0.56 to 0.86. The affinity computed via free energy perturbation (FEP) simulations is overestimated over the experimental value. Particularly, the electrostatic interaction free energy rules the binding process of ligands to AmpC and GluK1. However, the van der Waals (vdW) interaction free energy plays an important role in the ligand-binding processes of HSP90 and SARS-CoV-2 Mpro. The obtained results associate with the hydrophilic or hydrophobic properties of the ligands. This observation may enhance computer-aided drug design.

Morphology of a Transmembrane Aß42 Tetramer via REMD Simulations.

The folding/misfolding of membrane-permiable Amyloid beta (Aß) peptides is likely associated with the advancing stage of Alzheimer's disease (AD) by disrupting Ca2+ homeostasis. In this context, the aggregation of four transmembrane Aß17-42 peptides was investigated using temperature replica-exchange molecular dynamics (REMD) simulations. The obtained results indicated that the secondary structure of transmembrane Aß peptides tends to have different propensities compared to those in solution. Interestingly, the residues favorably forming ß-structure were interleaved by residues rigidly adopting turn-structure. A combination of ß and turn regions likely forms a pore structure. Six morphologies of 4Aß were found over the free energy landscape and clustering analyses. Among these, the morphologies include (1) Aß binding onto the membrane surface and three transmembrane Aß; (2) three helical and coil transmembrane Aß; (3) four helical transmembrane Aß; (4) three helical and one ß-hairpin transmembrane Aß; (5) two helical and two ß-strand transmembrane Aß; and (6) three ß-strand and one helical transmembrane Aß. Although the formation of the ß-barrel structure was not observed during the 0.28 ms─long MD simulation, the structure is likely to form when the simulation time is further extended.

Upgrading nirmatrelvir to inhibit SARS-CoV-2 Mpro via DeepFrag and free energy calculations.

The first oral drug for the treatment of COVID-19, Paxlovid, has been authorized; however, nirmatrelvir, a major component of the drug, is reported to be associated with some side effects. Moreover, the appearance of many novel variants raises concerns about drug resistance, and designing new potent inhibitors to prevent viral replication is thus urgent. In this context, using a hybrid approach combining machine learning (ML) and free energy simulations, 6 compounds obtained by modifying nirmatrelvir were proposed to bind strongly to SARS-CoV-2 Mpro. The structural modification of nirmatrelvir significantly enhances the electrostatic interaction free energy between the protein and ligand and slightly decreases the vdW term. However, the vdW term is the most important factor in controlling the ligand-binding affinity. In addition, the modified nirmatrelvir might be less toxic to the human body than the original inhibitor.

Machine learning combines atomistic simulations to predict SARS-CoV-2 Mpro inhibitors from natural compounds.

To date, the COVID-19 pandemic has still been infectious around the world, continuously causing social and economic damage on a global scale. One of the most important therapeutic targets for the treatment of COVID-19 is the main protease (Mpro) of SARS-CoV-2. In this study, we combined machine-learning (ML) model with atomistic simulations to computationally search for highly promising SARS-CoV-2 Mpro inhibitors from the representative natural compounds of the National Cancer Institute (NCI) Database. First, the trained ML model was used to scan the library quickly and reliably for possible Mpro inhibitors. The ML output was then confirmed using atomistic simulations integrating molecular docking and molecular dynamic simulations with the linear interaction energy scheme. The results turned out to show that there was evidently good agreement between ML and atomistic simulations. Ten substances were proposed to be able to inhibit SARS-CoV-2 Mpro. Seven of them have high-nanomolar affinity and are very potential inhibitors. The strategy has been proven to be reliable and appropriate for fast prediction of SARS-CoV-2 Mpro inhibitors, benefiting for new emerging SARS-CoV-2 variants in the future accordingly.

Searching for potential inhibitors of SARS-COV-2 main protease using supervised learning and perturbation calculations.

Inhibiting the biological activity of SARS-CoV-2 Mpro can prevent viral replication. In this context, a hybrid approach using knowledge- and physics-based methods was proposed to characterize potential inhibitors for SARS-CoV-2 Mpro. Initially, supervised machine learning (ML) models were trained to predict a ligand-binding affinity of ca. 2 million compounds with the correlation on a test set of R = 0.748 ± 0.044 . Atomistic simulations were then used to refine the outcome of the ML model. Using LIE/FEP calculations, nine compounds from the top 100 ML inhibitors were suggested to bind well to the protease with the domination of van der Waals interactions. Furthermore, the binding affinity of these compounds is also higher than that of nirmatrelvir, which was recently approved by the US FDA to treat COVID-19. In addition, the ligands altered the catalytic triad Cys145 - His41 - Asp187, possibly disturbing the biological activity of SARS-CoV-2.

Correction: Characterizing the ligand-binding affinity toward SARS-CoV-2 Mpro via physics- and knowledge-based approaches.

Correction for 'Characterizing the ligand-binding affinity toward SARS-CoV-2 Mpro via physics- and knowledge-based approaches' by Son Tung Ngo et al., Phys. Chem. Chem. Phys., 2022, https://doi.org/10.1039/d2cp04476e.

Correction: Rapid prediction of possible inhibitors for SARS-CoV-2 main protease using docking and FPL simulations.

[This corrects the article DOI: 10.1039/D0RA06212J.].

Characterizing the ligand-binding affinity toward SARS-CoV-2 Mpro via physics- and knowledge-based approaches.

Computational approaches, including physics- and knowledge-based methods, have commonly been used to determine the ligand-binding affinity toward SARS-CoV-2 main protease (Mpro or 3CLpro). Strong binding ligands can thus be suggested as potential inhibitors for blocking the biological activity of the protease. In this context, this paper aims to provide a short review of computational approaches that have recently been applied in the search for inhibitor candidates of Mpro. In particular, molecular docking and molecular dynamics (MD) simulations are usually combined to predict the binding affinity of thousands of compounds. Quantitative structure-activity relationship (QSAR) is the least computationally demanding and therefore can be used for large chemical collections of ligands. However, its accuracy may not be high. Moreover, the quantum mechanics/molecular mechanics (QM/MM) method is most commonly used for covalently binding inhibitors, which also play an important role in inhibiting the activity of SARS-CoV-2. Furthermore, machine learning (ML) models can significantly increase the searching space of ligands with high accuracy for binding affinity prediction. Physical insights into the binding process can then be confirmed via physics-based calculations. Integration of ML models into computational chemistry provides many more benefits and can lead to new therapies sooner.

Bayesian regression and model selection for isothermal titration calorimetry with enantiomeric mixtures.

Bayesian regression is performed to infer parameters of thermodynamic binding models from isothermal titration calorimetry measurements in which the titrant is an enantiomeric mixture. For some measurements the posterior density is multimodal, indicating that additional data with a different protocol are required to uniquely determine the parameters. Models of increasing complexity-two-component binding, racemic mixture, and enantiomeric mixture-are compared using model selection criteria. To precisely estimate one of these criteria, the Bayes factor, a variation of bridge sampling is developed.

Identifying Possible AChE Inhibitors from Drug-like Molecules via Machine Learning and Experimental Studies.

Acetylcholinesterase (AChE) is one of the most important drug targets for Alzheimer's disease (AD) treatment. In this work, a machine learning model was trained to rapidly and accurately screen large chemical databases for the potential inhibitors of AChE. The obtained results were then validated via in vitro enzyme assay. Moreover, atomistic simulations including molecular docking and molecular dynamics simulations were then used to understand molecular insights into the binding process of ligands to AChE. In particular, two compounds including benzyl trifluoromethyl ketone and trifluoromethylstyryl ketone were indicated as highly potent inhibitors of AChE because they established IC50 values of 0.51 and 0.33 µM, respectively. The obtained IC50 of two compounds is significantly lower than that of galantamine (2.10 µM). The predicted log(BB) suggests that the compounds may be able to traverse the blood-brain barrier. A good agreement between computational and experimental studies was observed, indicating that the hybrid approach can enhance AD therapy.

Searching for AChE inhibitors from natural compounds by using machine learning and atomistic simulations.

Acetylcholinesterase (AChE) is one of the most important drug targets for Alzheimer's disease treatment. In this work, a combined approach involving machine-learning (ML) model and atomistic simulations was established to predict the ligand-binding affinity to AChE of the natural compounds from VIETHERB database. The trained ML model was first utilized to rapidly and accurately screen the natural compound database for potential AChE inhibitors. Atomistic simulations including molecular docking and steered-molecular dynamics simulations were then used to confirm the ML outcome. Good agreement between ML and atomistic simulations was observed. Twenty compounds were suggested to be able to inhibit AChE. Especially, four of them including geranylgeranyl diphosphate, 2-phosphoglyceric acid, and 2-carboxy-d-arabinitol 1-phosphate, and farnesyl diphosphate are highly potent inhibitors with sub-nanomolar affinities.

Insights into the binding and covalent inhibition mechanism of PF-07321332 to SARS-CoV-2 Mpro.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been causing the COVID-19 pandemic, resulting in several million deaths being reported. Numerous investigations have been carried out to discover a compound that can inhibit the biological activity of the SARS-CoV-2 main protease, which is an enzyme related to the viral replication. Among these, PF-07321332 (Nirmatrelvir) is currently under clinical trials for COVID-19 therapy. Therefore, in this work, atomistic and electronic simulations were performed to unravel the binding and covalent inhibition mechanism of the compound to Mpro. Initially, 5 µs of steered-molecular dynamics simulations were carried out to evaluate the ligand-binding process to SARS-CoV-2 Mpro. The successfully generated bound state between the two molecules showed the important role of the PF-07321332 pyrrolidinyl group and the residues Glu166 and Gln189 in the ligand-binding process. Moreover, from the MD-refined structure, quantum mechanics/molecular mechanics (QM/MM) calculations were carried out to unravel the reaction mechanism for the formation of the thioimidate product from SARS-CoV-2 Mpro and the PF-07321332 inhibitor. We found that the catalytic triad Cys145-His41-Asp187 of SARS-CoV-2 Mpro plays an important role in the activation of the PF-07321332 covalent inhibitor, which renders the deprotonation of Cys145 and, thus, facilitates further reaction. Our results are definitely beneficial for a better understanding of the inhibition mechanism and designing new effective inhibitors for SARS-CoV-2 Mpro.

Effect of Cholesterol Molecules on Aß1-42 Wild-Type and Mutants Trimers.

Alzheimer's disease displays aggregates of the amyloid-beta (Aß) peptide in the brain, and there is increasing evidence that cholesterol may contribute to the pathogenesis of the disease. Though many experimental and theoretical studies have focused on the interactions of Aß oligomers with membrane models containing cholesterol, an understanding of the effect of free cholesterol on small Aß42 oligomers is not fully established. To address this question, we report on replica exchange with a solute tempering simulation of an Aß42 trimer with cholesterol and compare it with a previous replica exchange molecular dynamics simulation. We show that the binding hot spots of cholesterol are rather complex, involving hydrophobic residues L17-F20 and L30-M35 with a non-negligible contribution of loop residues D22-K28 and N-terminus residues. We also examine the effects of cholesterol on the trimers of the disease-causing A21G and disease-protective A2T mutations by molecular dynamics simulations. We show that these two mutations moderately impact cholesterol-binding modes. In our REST2 simulations, we find that cholesterol is rarely inserted into aggregates but rather attached as dimers and trimers at the surface of Aß42 oligomers. We propose that cholesterol acts as a glue to speed up the formation of larger aggregates; this provides a mechanistic link between cholesterol and Alzheimer's disease.

Unbinding ligands from SARS-CoV-2 Mpro via umbrella sampling simulations.

The umbrella sampling (US) simulation is demonstrated to be an efficient approach for determining the unbinding pathway and binding affinity to the SARS-CoV-2 Mpro of small molecule inhibitors. The accuracy of US is in the same range as the linear interaction energy (LIE) and fast pulling of ligand (FPL) methods. In detail, the correlation coefficient between US and experiments does not differ from FPL and is slightly smaller than LIE. The root mean square error of US simulations is smaller than that of LIE. Moreover, US is better than FPL and poorer than LIE in classifying SARS-CoV-2 Mpro inhibitors owing to the reciever operating characteristic-area under the curve analysis. Furthermore, the US simulations also provide detailed insights on unbinding pathways of ligands from the binding cleft of SARS-CoV-2 Mpro. The residues Cys44, Thr45, Ser46, Leu141, Asn142, Gly143, Glu166, Leu167, Pro168, Ala191, Gln192 and Ala193 probably play an important role in the ligand dissociation. Therefore, substitutions at these points may change the mechanism of binding of inhibitors to SARS-CoV-2 Mpro.

Improving ligand-ranking of AutoDock Vina by changing the empirical parameters.

AutoDock Vina (Vina) achieved a very high docking-success rate, p^ , but give a rather low correlation coefficient, R , for binding affinity with respect to experiments. This low correlation can be an obstacle for ranking of ligand-binding affinity, which is the main objective of docking simulations. In this context, we evaluated the dependence of Vina R coefficient upon its empirical parameters. R is affected more by changing the gauss2 and rotation than other terms. The docking-success rate p^ is sensitive to the alterations of the gauss1, gauss2, repulsion, and hydrogen bond parameters. Based on our benchmarks, the parameter set1 has been suggested to be the most optimal. The testing study over 800 complexes indicated that the modified Vina provided higher correlation with experiment Rset1=0.556±0.025 compared with RDefault=0.493±0.028 obtained by the original Vina and RVina1.2=0.503±0.029 by Vina version 1.2. Besides, the modified Vina can be also applied more widely, giving R≥0.500 for 32/48 targets, compared with the default package, giving R≥0.500 for 31/48 targets. In addition, validation calculations for 1036 complexes obtained from version 2019 of PDBbind refined structures showed that the set1 of parameters gave higher correlation coefficient ( Rset1=0.617±0.017 ) than the default package ( RDefault=0.543±0.020 ) and Vina version 1.2 ( RVina1.2=0.540±0.020 ). The version of Vina with set1 of parameters can be downloaded at https://github.com/sontungngo/mvina. The outcomes would enhance the ranking of ligand-binding affinity using Autodock Vina.

Benchmark of Popular Free Energy Approaches Revealing the Inhibitors Binding to SARS-CoV-2 Mpro.

The COVID-19 pandemic has killed millions of people worldwide since its outbreak in December 2019. The pandemic is caused by the SARS-CoV-2 virus whose main protease (Mpro) is a promising drug target since it plays a key role in viral proliferation and replication. Currently, developing an effective therapy is an urgent task, which requires accurately estimating the ligand-binding free energy to SARS-CoV-2 Mpro. However, it should be noted that the accuracy of a free energy method probably depends on the protein target. A highly accurate approach for some targets may fail to produce a reasonable correlation with the experiment when a novel enzyme is considered as a drug target. Therefore, in this context, the ligand-binding affinity to SARS-CoV-2 Mpro was calculated via various approaches. The molecular docking approach was manipulated using Autodock Vina (Vina) and Autodock4 (AD4) protocols to preliminarily investigate the ligand-binding affinity and pose to SARS-CoV-2 Mpro. The binding free energy was then refined using the fast pulling of ligand (FPL), linear interaction energy (LIE), molecular mechanics-Poisson-Boltzmann surface area (MM-PBSA), and free energy perturbation (FEP) methods. The benchmark results indicated that for docking calculations, Vina is more accurate than AD4, and for free energy methods, FEP is the most accurate method, followed by LIE, FPL, and MM-PBSA (FEP > LIE ≈ FPL > MM-PBSA). Moreover, atomistic simulations revealed that the van der Waals interaction is the dominant factor. The residues Thr26, His41, Ser46, Asn142, Gly143, Cys145, His164, Glu166, and Gln189 are essential elements affecting the binding process. Our benchmark provides guidelines for further investigations using computational approaches.

Thermodynamics and kinetics in antibody resistance of the 501Y.V2 SARS-CoV-2 variant.

Understanding the thermodynamics and kinetics of the binding process of an antibody to the SARS-CoV-2 receptor-binding domain (RBD) of the spike protein is very important for the development of COVID-19 vaccines. In particular, it is essential to understand how the binding mechanism may change under the effects of RBD mutations. In this context, we have demonstrated that the South African variant (B1.351 or 501Y.V2) can resist the neutralizing antibody (NAb). Three substitutions in the RBD including K417N, E484K, and N501Y alter the free energy landscape, binding pose, binding free energy, binding kinetics, hydrogen bonding, nonbonded contacts, and unbinding pathway of RBD + NAb complexes. The low binding affinity of NAb to 501Y.V2 RBD confirms the antibody resistance of the South African variant. Moreover, the fragment of NAb + RBD can be used as an affordable model to investigate changes in the binding process between the mutated RBD and antibodies.

Implicit ligand theory for relative binding free energies: II. An estimator based on control variates.

Implicit ligand theory describes the relationship between the noncovalent binding free energy and the binding free energy between a ligand and multiple rigid receptor conformations. We have previously shown that if the receptor conformations are sampled from or reweighed to a holo ensemble, the binding free energy relative to the ligand that defines the ensemble can be calculated. Here, we apply a variance reduction technique known as control variates to derive a new statistical estimator for the relative binding free energy. In applications to a data set of 6 reference ligands and 18 test ligands, statistically significant differences between the estimators are not observed for most systems. However, in cases where such differences are observed, the new estimator is more accurate, precise, and converges more quickly. Performance improvements are most consistent where there is a clear correlation, with a correlation coefficient greater than 0.3, between the control variate and the statistic being averaged.

Rapid prediction of possible inhibitors for SARS-CoV-2 main protease using docking and FPL simulations.

Originating for the first time in Wuhan, China, the outbreak of SARS-CoV-2 has caused a serious global health issue. An effective treatment for SARS-CoV-2 is still unavailable. Therefore, in this study, we have tried to predict a list of potential inhibitors for SARS-CoV-2 main protease (Mpro) using a combination of molecular docking and fast pulling of ligand (FPL) simulations. The approaches were initially validated over a set of eleven available inhibitors. Both Autodock Vina and FPL calculations produced consistent results with the experiments with correlation coefficients of R Dock = 0.72 ± 0.14 and R W = -0.76 ± 0.10, respectively. The combined approaches were then utilized to predict possible inhibitors that were selected from a ZINC15 sub-database for SARS-CoV-2 Mpro. Twenty compounds were suggested to be able to bind well to SARS-CoV-2 Mpro. Among them, five top-leads are periandrin V, penimocycline, cis-p-Coumaroylcorosolic acid, glycyrrhizin, and uralsaponin B. The obtained results could probably lead to enhance the COVID-19 therapy.

Oversampling Free Energy Perturbation Simulation in Determination of the Ligand-Binding Free Energy.

Determination of the ligand-binding affinity is an extremely interesting problem. Normally, the free energy perturbation (FEP) method provides an appropriate result. However, it is of great interest to improve the accuracy and precision of this method. In this context, temperature replica exchange molecular dynamics implementation of the FEP computational approach, which we call replica exchange free energy perturbation (REP) was proposed. In particular, during REP simulations, the system can easily escape from being trapped in local minima by exchanging configurations with high temperatures, resulting in significant improvement in the accuracy and precision of protein-ligand binding affinity calculations. The distribution of the decoupling free energy was enlarged, and its mean values were decreased. This results in changes in the magnitude of the calculated binding free energies as well as in alteration in the binding mechanism. Moreover, the REP correlation coefficient with respect to experiment ( RREP = 0.85 ± 0.15) is significantly boosted in comparison with the FEP one ( RFEP = 0.64 ± 0.30). Furthermore, the root-mean-square error (RMSE) of REP is also smaller than FEP, RMSEREP = 4.28 ± 0.69 versus RMSEFEP = 5.80 ± 1.11 kcal/mol, respectively. © 2019 Wiley Periodicals, Inc.