Robust optimization of scoring functions for a target class

Target-specific optimization of scoring functions for protein–ligand docking is an effective method for significantly improving the discrimination of active and inactive molecules in virtual screening applications. Its applicability, however, is limited due to the narrow focus on, e.g., single protein structures. Using an ensemble of protein kinase structures, the publically available directory of useful decoys ligand dataset, and a novel multi-factorial optimization procedure, it is shown here that scoring functions can be tuned to multiple targets of a target class simultaneously. This leads to an improved robustness of the resulting scoring function parameters. Extensive validation experiments clearly demonstrate that (1) virtual screening performance for kinases improves significantly; (2) variations in database content affect this kind of machine-learning strategy to a lesser extent than binary QSAR models, and (3) the reweighting of interaction types is of particular importance for improved screening performance. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

A fast method of large-scale serial semiempirical calculations of docking complexes

A brief survey of the state of the art in methods of calculations of protein—ligand interaction energies in docking complexes is presented. A new computational technique is proposed that allows one to fundamentally improve the performance of large-scale serial calculations of docking complexes using the AM1/PM3 semiempirical methods. The technique explicitly allows for a specific feature of docking problems, viz., the need for calculating numerous ligand complexes with a specified protein whose noninteracting part remains “frozen” during computations. The interaction energies calculated using the new method differ only slightly from the results of complete AM1 calculations and the performance attained is high enough to solve practical drug design problems. Published in Russian in Izvestiya Akademii Nauk. Seriya Khimicheskaya, No. 9, pp. 1759–1764, September, 2008.     

Noise reduction method for molecular interaction energy: application to in silico drug screening and in silico target protein screening

We developed a new method to improve the accuracy of molecular interaction data using a molecular interaction matrix. This method was applied to enhance the database enrichment of in silico drug screening and in silico target protein screening using a protein-compound affinity matrix calculated by a protein-compound docking software. Our assumption was that the protein-compound binding free energy of a compound could be improved by a linear combination of its docking scores with many different proteins. We proposed two approaches to determine the coefficients of the linear combination. The first approach is based on similarity among the proteins, and the second is a machine-learning approach based on the known active compounds. These methods were applied to in silico screening of the active compounds of several target proteins and in silico target protein screening.     

Studying protein–protein affinity and immobilized ligand–protein affinity interactions using MS-based methods

This review discusses the most important current methods employing mass spectrometry (MS) analysis for the study of protein affinity interactions. The methods are discussed in depth with particular reference to MS-based approaches for analyzing protein–protein and protein–immobilized ligand interactions, analyzed either directly or indirectly. First, we introduce MS methods for the study of intact protein complexes in the gas phase. Next, pull-down methods for affinity-based analysis of protein–protein and protein–immobilized ligand interactions are discussed. Presently, this field of research is often called interactomics or interaction proteomics. A slightly different approach that will be discussed, chemical proteomics, allows one to analyze selectivity profiles of ligands for multiple drug targets and off-targets. Additionally, of particular interest is the use of surface plasmon resonance technologies coupled with MS for the study of protein interactions. The review addresses the principle of each of the methods with a focus on recent developments and the applicability to lead compound generation in drug discovery as well as the elucidation of protein interactions involved in cellular processes. The review focuses on the analysis of bioaffinity interactions of proteins with other proteins and with ligands, where the proteins are considered as the bioactives analyzed by MS.     

Stable Isotope Labeling Strategy for Protein–Ligand Binding Analysis in Multi-Component Protein Mixtures

Described here is a stable isotope labeling protocol that can be used with a chemical modification- and mass spectrometry-based protein–ligand binding assay for detecting and quantifying both the direct and indirect binding events that result from protein–ligand binding interactions. The protocol utilizes an H₂¹⁶O₂ and H₂¹⁸O₂ labeling strategy to evaluate the chemical denaturant dependence of methionine oxidation in proteins both in the presence and absence of a target ligand. The differential denaturant dependence to the oxidation reactions performed in the presence and absence of ligand provides a measure of the protein stability changes that occur as a result of direct interactions of proteins with the target ligand and/or as a result of indirect interactions involving other protein–ligand interactions that are either induced or disrupted by the ligand. The described protocol utilizes the ¹⁸O/¹⁶O ratio in the oxidized protein samples to quantify the ligand-induced protein stability changes. The ratio is determined using the isotopic distributions observed for the methionine-containing peptides used for protein identification in the LC-MS-based proteomics readout. The strategy is applied to a multi-component protein mixture in this proof-of-principle experiment, which was designed to evaluate the technique’s ability to detect and quantify the direct binding interaction between cyclosporin A and cyclophilin A and to detect the indirect binding interaction between cyclosporin A and calcineurin (i.e., the protein–protein interaction between cyclophilin A and calcineurin that is induced by cyclosporin A binding to cyclophilin A).     

Fourier transform infrared spectroscopy as a tool to study structural properties of cytochromes P450 (CYPs)

Cytochrome P450 proteins (CYPs) are a big class of heme proteins which are involved in various metabolic processes of living organisms. CYPs are the terminal catalytically active components of monooxygenase systems where the substrate binds and is hydroxylated. In order to be functionally competent, the protein structures of CYPs possess specific properties that must be explored in order to understand structure–function relationships and mechanistic aspects. Fourier transform infrared spectroscopy (FTIR) is one tool that is used to study these structural properties. The application of FTIR spectroscopy to the secondary structures of CYP proteins, protein unfolding, protein–protein interactions and the structure and dynamics of the CYP heme pocket is reviewed. A comparison with other thiolate heme proteins (nitric oxide synthase and chloroperoxidase) is also included. Figure The protein secondary structure, protein unfolding, redox-partner protein–protein interaction, structural changes induced by the reduction of the heme iron, and the structure and dynamics of the active site of cytochromes P450 (CYP) can be studied using Fourier transform infrared spectroscopy (FTIR). FTIR spectroscopy is a good approach for gaining a deeper insight into structure–function relationships in CYPs. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Virtual Screening with Gnina 1.0

Virtual screening—predicting which compounds within a specified compound library bind to a target molecule, typically a protein—is a fundamental task in the field of drug discovery. Doing virtual screening well provides tangible practical benefits, including reduced drug development costs, faster time to therapeutic viability, and fewer unforeseen side effects. As with most applied computational tasks, the algorithms currently used to perform virtual screening feature inherent tradeoffs between speed and accuracy. Furthermore, even theoretically rigorous, computationally intensive methods may fail to account for important effects relevant to whether a given compound will ultimately be usable as a drug. Here we investigate the virtual screening performance of the recently released Gnina molecular docking software, which uses deep convolutional networks to score protein-ligand structures. We find, on average, that Gnina outperforms conventional empirical scoring. The default scoring in Gnina outperforms the empirical AutoDock Vina scoring function on 89 of the 117 targets of the DUD-E and LIT-PCBA virtual screening benchmarks with a median 1% early enrichment factor that is more than twice that of Vina. However, we also find that issues of bias linger in these sets, even when not used directly to train models, and this bias obfuscates to what extent machine learning models are achieving their performance through a sophisticated interpretation of molecular interactions versus fitting to non-informative simplistic property distributions.     

Gp41 inhibitory activity prediction of theaflavin derivatives using ligand/structure-based virtual screening approaches

Gp41 and its conserved hydrophobic groove on the NHR region is one of the attractive targets in the design of HIV-1 entry inhibitory agents. This hydrophobic pocket is very critical for the progression of HIV and host cell fusion. In this study different ligand-based (structure similarity search) and structure-based (molecular docking and molecular dynamic simulation) methods were performed in a virtual screening procedure to select the best compounds with the most probable HIV-1 gp41 inhibitory activities. In silico pharmacokinetics and ADMET (absorption, distribution, metabolism, excretion and toxicity) properties filtration also was considered to choose the compounds with best drug-like properties. The results of molecular docking and molecular dynamic simulations of the final selected compounds showed suitable stabilities of their complexes with gp41. The final selected hits could have better pharmacokinetics properties than the template compound, theaflavin digallate (TF₃), a naturally-originated potent gp41 inhibitor.     

A new peptide docking strategy using a mean field technique with mutually orthogonal Latin square sampling

The theoretical prediction of the association of a flexible ligand with a protein receptor requires efficient sampling of the conformational space of the ligand. Several docking methodologies are currently available. We propose a new docking technique that performs well at low computational cost. The method uses mutually orthogonal Latin squares to efficiently sample the docking space. A variant of the mean field technique is used to analyze this sample to arrive at the optimum. The method has been previously applied to explore the conformational space of peptides and identify structures with low values for the potential energy. Here we extend this method to simultaneously identify both the low energy conformation as well as a ‘high-scoring’ docking mode. Application of the method to 56 protein–peptide complexes, in which the length of the peptide ligand ranges from three to seven residues, and comparisons with Autodock 3.05, showed that the method works well. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Supercomputer docking with a large number of degrees of freedom

ABSTRACT

Docking represents one of the most popular computational approaches in drug design. It has reached popularity owing to capability of identifying correct conformations of a ligand within an active site of the target-protein and of estimating the binding affinity of a ligand that is immensely helpful in prediction of compound activity. Despite many success stories, there are challenges, in particular, handling with a large number of degrees of freedom in solving the docking problem. Here, we show that SOL-P, the docking program based on the new Tensor Train algorithm, is capable to dock successfully oligopeptides having up to 25 torsions. To make the study comparative we have performed docking of the same oligopeptides with the SOL program which uses the same force field as that utilized by SOL-P and has common features of many docking programs: the genetic algorithm of the global optimization and the grid approximation. SOL has managed to dock only one oligopeptide. Moreover, we present the results of docking with SOL-P ligands into proteins with moveable atoms. Relying on visual observations we have determined the common protein atom groups displaced after docking which seem to be crucial for successful prediction of experimental conformations of ligands.     

Docking-based virtual screening studies aiming at the covalent inhibition of SARS-CoV-2 MPro by targeting the cysteine 145

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for the COVID-19 which has infected millions of people worldwide. The main protease of SARS-CoV-2 (M^Pro) has been recognized as a key target for the development of antiviral compounds. Taking advantage of the X-ray crystal complex with reversible covalent inhibitors interacting with the catalytic cysteine 145 (Cys145), we explored flexible docking studies to select alternative compounds able to target this residue as covalent inhibitors. First, docking studies of three known electrophilic compounds led to results consistent with co-crystallized data validating the method for SARS-CoV-2 M^Pro covalent inhibition. Then, libraries of soft electrophiles (overall 41 757 compounds) were submitted to docking-based virtual screening resulting in the identification of 17 molecules having their electrophilic group close to the Cys145 residue. We also investigated flexible docking studies of a focused approved covalent drugs library including 32 compounds with various electrophilic functional groups. Among them, the calculations resulted in the identification of four compounds, namely dimethylfumarate, fosfomycin, ibrutinib and saxagliptin, able first, to bind to the active site of the protein and second, to form a covalent bond with the catalytic cysteine.     

Modeling the Dynamics of Protein–Protein Interfaces,How and Why?

Protein–protein assemblies act as a key component in numerous cellular processes. Their accurate modeling at the atomic level remains a challenge for structural biology. To address this challenge, several docking and a handful of deep learning methodologies focus on modeling protein–protein interfaces. Although the outcome of these methods has been assessed using static reference structures, more and more data point to the fact that the interaction stability and specificity is encoded in the dynamics of these interfaces. Therefore, this dynamics information must be taken into account when modeling and assessing protein interactions at the atomistic scale. Expanding on this, our review initially focuses on the recent computational strategies aiming at investigating protein–protein interfaces in a dynamic fashion using enhanced sampling, multi-scale modeling, and experimental data integration. Then, we discuss how interface dynamics report on the function of protein assemblies in globular complexes, in fuzzy complexes containing intrinsically disordered proteins, as well as in active complexes, where chemical reactions take place across the protein–protein interface.     

Virtual Screening of Human O-GlcNAc Transferase Inhibitors

O-GlcNAc transferase (OGT) is one of essential mammalian enzymes, which catalyze the transfer of N-acetylglucosamine from UDP-N-acetylglucosamine (UDP-GlcNAc) to hydroxyl groups of serines and threonines (Ser/Thr) in proteins. Dysregulations of cellular O-GlcNAc have been implicated in diabetes, neurodegenerative disease, and cancer, which brings great interest in developing potent and speci c small-molecular OGT inhibitors. In this work, we performed virtual screening on OGT catalytic site to identify potential inhibitors. 7134792 drug-like compounds from ZINC (a free database of commercially available compounds for virtual screening) and 4287550 compounds generated by FOG (fragment optimized growth program) were screened and the top 116 compounds ranked by docking score were analyzed. By comparing the screening results, we found FOG program can generate more compounds with better docking scores than ZINC. The top ZINC compounds ranked by docking score were grouped into two classes, which held the binding positions of UDP and GlcNAc of UDPGlcNAc. Combined with individual fragments in binding pocket, de novo compounds were designed and proved to have better docking score. The screened and designed compounds may become a starting point for developing new drugs.     

Homology modeling and molecular dynamics simulation of N-myristoyltransferase from protozoan parasites: active site characterization and insights into rational inhibitor design

Myristoyl-CoA:protein N-myristoyltransferase (NMT) is a cytosolic monomeric enzyme that catalyzes the transfer of the myristoyl group from myristoyl-CoA to the N-terminal glycine of a number of eukaryotic cellular and viral proteins. Recent experimental data suggest NMT from parasites could be a promising new target for the design of novel antiparasitic agents with new mode of action. However, the active site topology and inhibitor specificity of these enzymes remain unclear. In this study, three-dimensional models of NMT from Plasmodium falciparum (PfNMT), Leishmania major (LmNMT) and Trypanosoma brucei (TbNMT) were constructed on the basis of the crystal structures of fungal NMTs using homology modeling method. The models were further refined by energy minimization and molecular dynamics simulations. The active sites of PfNMT, LmNMT and TbNMT were characterized by multiple copy simultaneous search (MCSS). MCSS functional maps reveal that PfNMT, LmNMT and TbNMT share a similar active site topology, which is defined by two hydrophobic pockets, a hydrogen-bonding (HB) pocket, a negatively-charged HB pocket and a positively-charged HB pocket. Flexible docking approaches were then employed to dock known inhibitors into the active site of PfNMT. The binding mode, structure–activity relationships and selectivity of inhibitors were investigated in detail. From the results of molecular modeling, the active site architecture and certain key residues responsible for inhibitor binding were identified, which provided insights for the design of novel inhibitors of parasitic NMTs.     

Computational study of the heterodimerization between μ and δ receptors

A growing body of evidence indicated that the G protein coupled receptors exist as homo- or hetero-dimers in the living cell. The heterodimerization between μ and δ opioid receptors has attracted researchers’ particular interests, it is reported to display novel pharmacological and signalling regulation properties. In this study, we construct the full-length 3D-model of μ and δ opioid receptors using the homology modelling method. Threading program was used to predict the possible templates for the N- and C-terminus domains. Then, a 30 ns molecular dynamics simulations was performed with each receptor embedded in an explicit membrane-water environment to refine and explore the conformational space. Based on the structures extracted from the molecular dynamics, the likely interface of μ–δ heterodimer was investigated through the analysis of protein–protein docking, cluster, shape complementary and interaction energy. The computational modelling works revealed that the most likely interface of heterodimer was formed between the transmembrane1,7 (TM1,7) domains of μ receptor and the TM(4,5) domains of δ receptor, with emphasis on μ-TM1 and δ-TM4, the next likely interface was μ(TM6,7)-δ(TM4,5), with emphasis on μ-TM6 and δ-TM4. Our results were consistent with previous reports. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Preference Prediction for the Stable Inclusion Complexes between Cyclodextrins and Monocyclic Insoluble Chemicals Based on Monte Carlo Docking Simulations

Cyclodextrins (CDs) are useful functional excipients, which are being used to camouflage undesirable pharmaceutical characteristics, especially poor aqueous solubility, through the inclusion complexation process with insoluble drugs. The selection of more efficient cyclodextrin is important to improve the bioavailability of drugs. In this study, the complexing and solubilizing abilities toward poorly water-soluble monocyclic molecules of natural CDs (α-CD, β-CD, and γ-CD) were investigated using Monte Carlo (MC) docking simulations studies. These theoretical results closely agree with the experimental observation of the complex stability in water of the various guests–CD complexes. Host preferences, based on the experimentally determined stability constants between host CDs and guest molecules, show excellent correlation with the calculated interaction energies of corresponding complexes. The inclusion complex with the lower MC docking interaction energy shows a higher value of stability constant than that of the other complex, and the prediction accuracy of the preferred complex for 21 host–guest pairs is 100%. This result indicates that the MC docking interaction energy could be employed as a useful parameter to select more efficient cyclodextrin as a host for the bioavailability of insoluble drugs. In this study, β-CD shows greater solubilizing efficacies toward guest molecules than those of α-CD and γ-CD, with the exception of one case due to the structure of a guest molecule containing one lipophilic cyclic moiety. The surface area change of CDs and hydrogen bonding between the host and guest also work as major factors for the formation of the stable complex.     

Identification of potential Tpx inhibitors against pathogen-host interactions

Yersinia organisms cause many infectious diseases by invading human cells and delivering their virulence factors via the type three secretion system (T3SS). One alternative strategy in the fight against these pathogenic organisms is to interfere with their T3SS. Previous studies demonstrated that thiol peroxidase, Tpx is functional in the assembly of T3SS and its inhibition by salicylidene acylhydrazides prevents the secretion of pathogenic effectors. In this study, the aim was to identify potential inhibitors of Tpx using an integrated approach starting with high throughput virtual screening and ending with molecular dynamics simulations of selected ligands. Virtual screening of ZINC database of 500,000 compounds via ligand-based and structure-based pharmacophore models retrieved 10,000 hits. The structure-based pharmacophore model was validated using high-throughput virtual screening (HTVS). After multistep docking (SP and XP), common scaffolds were used to find common substructures and the ligand binding poses were optimized using induced fit docking. The stability of the protein–ligand complex was examined with molecular dynamics simulations and the binding free energy of the complex was calculated. As a final outcome eight compounds with different chemotypes were proposed as potential inhibitors for Tpx. The eight ligands identified by a detailed virtual screening protocol can serve as leads in future drug design efforts against the destructive actions of pathogenic bacteria.