Collection and preparation of molecular databases for virtual screening

Drug discovery and development research is undergoing a paradigm shift from a linear and sequential nature of the various steps involved in the drug discovery process of the past to the more parallel approach of the present, due to a lack of sufficient correlation between activities estimated by in vitro and in vivo assays. This is attributed to the non-drug-likeness of the lead molecules, which has often been detected at advanced drug development stages. Thus a striking aspect of this paradigm shift has been early/parallel in silico prioritization of drug-like molecular databases (also database pre-processing), in addition to prioritizing compounds with high affinity and selectivity for a protein target. In view of this, a drug-like database useful for virtual screening has been created by prioritizing molecules from 36 catalog suppliers, using our recently derived binary QSAR based drug-likeness model as a filter. The performance of this model was assessed by a comparative evaluation with respect to commonly used filters implemented by the ZINC database. Since the model was derived considering all the limitations that have plagued the existing rules and models, it performs better than the existing filters and thus the molecules prioritized by this filter represent a better subset of drug-like compounds. The application of this model on exhaustive subsets of 4,972,123 molecules, many of which have passed the ZINC database filters for drug-likeness, led to a further prioritization of 2,920,551 drug-like molecules. This database may have a great potential for in silico virtual screening for discovering molecules, which may survive the later stages of the drug development research.     

Identification of potential Gly/NMDA receptor antagonists by cheminformatics approach: a combination of pharmacophore modelling,virtual screening and molecular docking studies

The Gly/NMDA receptor has become known as potential target for the management of neurodegenerative diseases. Discovery of Gly/NMDA antagonists has thus attracted much attention in recent years. In the present research, a cheminformatics approach has been used to determine structural requirements for Gly/NMDA antagonism and to identify potential antagonists. Here, 37 quinoxaline derivatives were selected to develop a significant pharmacophore model with good certainty. The selected model was validated by leave-one-out cross-validation, an external test set, decoy set and Y-randomization test. Applicability domain was verified by the standardization approach. The validated 3D-QSAR model was used to screen virtual hits from the ZINC database by pharmacophore mapping. Molecular docking was used for assessment of receptor–ligand binding modes and binding affinities. The GlideScore and molecular interactions with critical amino acids were considered as crucial features to identify final hits. Furthermore, hits were analysed for in silico pharmacokinetic parameters and Lipinski’s rule of five, demonstrating their potential as drug-like candidates. The PubChem and SciFinder search tools were used to authenticate the novelty of leads retrieved. Finally, five different leads have been suggested as putative novel candidates for the exploration of potent Gly/NMDA receptor antagonists.     

Pharmacophore modelling,molecular docking and virtual screening for EGFR (HER 1) tyrosine kinase inhibitors

A pharmacophore model has been developed using diverse classes of epidermal growth factor receptor (EGFR) tyrosine kinase (TK) inhibitors useful in the treatment of human tumours. Among the top 10 generated hypotheses, the second hypothesis, with one hydrogen bond acceptor, one ring aromatic and three hydrophobic features, was found to be the best on the basis of Cat Scramble validation as well as test set prediction (r _training?=?0.89, r _test?=?0.82). The model also maps well to the external test set molecules as well as clinically active molecules and corroborates the docking studies. Finally, 10 hits were identified as potential leads after virtual screening of ZINC database for EGFR TK inhibition. The study may facilitate the designing and discovery of novel EGFR TK inhibitors.     

Structure-guided fragment-based<Emphasis Type="Italic"> in silico</Emphasis> drug design of dengue protease inhibitors

An in silico fragment-based drug design approach was devised and applied towards the identification of small molecule inhibitors of the dengue virus (DENV) NS2B-NS3 protease. Currently, no DENV protease co-crystal structure with bound inhibitor and fully formed substrate binding site is available. Therefore a homology model of DENV NS2B-NS3 protease was generated employing a multiple template spatial restraints method and used for structure-based design. A library of molecular fragments was derived from the ZINC screening database with help of the retrosynthetic combinatorial analysis procedure (RECAP). 150,000 molecular fragments were docked to the DENV protease homology model and the docking poses were rescored using a target-specific scoring function. High scoring fragments were assembled to small molecule candidates by an implicit linking cascade. The cascade included substructure searching and structural filters focusing on interactions with the S1 and S2 pockets of the protease. The chemical space adjacent to the promising candidates was further explored by neighborhood searching. A total of 23 compounds were tested experimentally and two compounds were discovered to inhibit dengue protease (IC₅₀ = 7.7 μM and 37.9 μM, respectively) and the related West Nile virus protease (IC₅₀ = 6.3 μM and 39.0 μM, respectively). This study demonstrates the successful application of a structure-guided fragment-based in silico drug design approach for dengue protease inhibitors providing straightforward hit generation using a combination of homology modeling, fragment docking, chemical similarity and structural filters.     

Structure-based model profiles affinity constant of drugs with hPEPT1 for rapid virtual screening of hPEPT1’s substrate

The human proton-coupled peptide transporter (hPEPT1) with broad substrates is an important route for improving the pharmacokinetic performance of drugs. Thus, it is essential to predict the affinity constant between drug molecule and hPEPT1 for rapid virtual screening of hPEPT1’s substrate during lead optimization, candidate selection and hPEPT1 prodrug design. Here, a structure-based in silico model for 114 compounds was constructed based on eight structural parameters. This model was built by the multiple linear regression method and satisfied all the prerequisites of the regression models. For the entire data set, the r² and adjusted r² values were 0.74 and 0.72, respectively. Then, this model was used to perform substrate/non-substrate classification. For 29 drugs from DrugBank database, all were correctly classified as substrates of hPEPT1. This model was also used to perform substrate/non-substrate classification for 18 drugs and their prodrugs; this QSAR model also can distinguish between the substrate and non-substrate. In conclusion, the QSAR model in this paper was validated by a large external data set, and all results indicated that the developed model was robust, stable, and can be used for rapid virtual screening of hPEPT1’s substrate in the early stage of drug discovery.     

Ligand based virtual screening using SVM on GPU

In silico methods play an essential role in modern drug discovery methods. Virtual screening, an in silico method, is used to filter out the chemical space on which actual wet lab experiments are need to be conducted. Ligand based virtual screening is a computational strategy using which one can build a model of the target protein based on the knowledge of the ligands that bind successfully to the target. This model is then used to predict if the new molecule is likely to bind to the target. Support vector machine, a supervised learning algorithm used for classification, can be utilized for virtual screening the ligand data. When used for virtual screening purpose, SVM could produce interesting results. But since we have a huge ligand data, the time taken for training the SVM model is quite high compared to other learning algorithms. By parallelizing these algorithms on multi-core processors, one can easily expedite these discoveries. In this paper, a GPU based ligand based virtual screening tool (GpuSVMScreen) which uses SVM have been proposed and bench-marked. This data parallel virtual screening tool provides high throughput by running in short time. The proposed GpuSVMScreen can successfully screen large number of molecules (billions) also. The source code of this tool is available at http://ccc.nitc.ac.in/project/GPUSVMSCREEN.     

In silico drug repositioning of FDA-approved drugs to predict new inhibitors for alpha-synuclein aggregation

One of the hallmarks of Parkinson’s disease (PD), a long-term neurodegenerative syndrome, is the accumulation of alpha-synuclein (α-syn) fibrils. Despite numerous studies and efforts, inhibition of α-syn protein aggregation is still a challenge. To overcome this issue, we propose an in silico pharmacophore-based repositioning strategy, to find a pharmaceutical drug that, in addition to their defined role, can be used to prevent aggregation of the α-syn protein. Ligand-based pharmacophore modeling was developed and the best model was selected with validation parameters including 72 % sensitivity, 98 % specificity and goodness score about 0.7. The optimal model has three groups of hydrogen bond donor (HBD), three groups of hydrogen bond acceptor (HBA), and two aromatic rings (AR). The FDA-Approved reports in the ZINC15 database were screened with the pharmacophore model taken from inhibitor compounds. The model identified 22 hits, as promising candidate drugs for Parkinson's therapy. It is noteworthy that among these, 10 drugs have been reported to inhibition of α-syn aggregation or treat/reduce Parkinson's pathogenesis. This model was used to virtual screen ZINC, NCI databases, and natural products from the pomegranate. The results of this screen were filtered for their inability to cross the blood-brain barrier, poor oral bioavailability, etc. Finally, the selected compounds of two ZINC and NCI databases were combined and structurally clustered. Remained compounds were clustered in 28 different clusters, and the 17 compounds were introduced as final candidates.     

Investigation of novel compounds via in silico approaches of EGFR inhibitors as anticancer agents

Before the experimental studies of a compound to be synthesized from in vitro to in vivo, it is possible to save both time and money with in silico approaches only with Computer Aided Drug Design (CADD) methods. In other words, compounds that can be new drug candidates can be suggested by drug design using computational drug discovery strategies. In this study, all molecules in the ChEMBL Database were virtual screened based on drugs with inhibitory properties on the Epidermal Growth Factor Receptor (EGFR), one of the receptor tyrosine kinases, which is effective in cancer cells. During this High-Throughput Screening (HTS), the number of compounds was minimized according to the parameters of the reference drugs, physicochemical properties such as logP, M.W., HBA, HDB, RosLip. As a result of in silico approaches and molecular docking analysis, ten compounds with the highest docking scores were determined and a model compound that could be a new drug candidate was proposed.     

Identification of novel Plasmodium falciparum PI4KB inhibitors as potential anti-malarial drugs: Homology modeling,molecular docking and molecular dynamics simulations

The current study was set to discover selective Plasmodium falciparum phosphatidylinositol-4-OH kinase type III beta (pfPI4KB) inhibitors as potential antimalarial agents using combined structure-based and ligand-based drug discovery approach. A comparative model of pfPI4KB was first constructed and validated using molecular docking techniques. Performance of Autodock4.2 and Vina4 software in predicting the inhibitor-PI4KB binding mode and energy was assessed based on two Test Sets: Test Set I contained five ligands with resolved crystal structures with PI4KB, while Test Set II considered eleven compounds with known IC50 value towards PI4KB. The outperformance of Autodock as compared to Vina was reported, giving a correlation coefficient (R²) value of 0.87 and 0.90 for Test Set I and Test Set II, respectively. Pharmacophore-based screening was then conducted to identify drug-like molecules from ZINC database with physicochemical similarity to two potent pfPI4KB inhibitors –namely cpa and cpb. For each query inhibitor, the best 1000 hits in terms of TanimotoCombo scores were selected and subjected to molecular docking and molecular dynamics (MD) calculations. Binding energy was then estimated using molecular mechanics–generalized Born surface area (MM-GBSA) approach over 50 ns MD simulations of the inhibitor-pfPI4KB complexes. According to the calculated MM-GBSA binding energies, ZINC78988474 and ZINC20564116 were identified as potent pfPI4KB inhibitors with binding energies better than those of cpa and cpb, with ΔG_binding ≥ −34.56 kcal/mol. The inhibitor-pfPI4KB interaction and stability were examined over 50 ns MD simulation; as well the selectivity of the identified inhibitors towards pfPI4KB over PI4KB was reported.     

Machine-Learning-Enabled Virtual Screening for Inhibitors of Lysine-Specific Histone Demethylase 1

A machine learning approach has been applied to virtual screening for lysine specific demethylase 1 (LSD1) inhibitors. LSD1 is an important anti-cancer target. Machine learning models to predict activity were constructed using Morgan molecular fingerprints. The dataset, consisting of 931 molecules with LSD1 inhibition activity, was obtained from the ChEMBL database. An evaluation of several candidate algorithms on the main dataset revealed that the support vector regressor gave the best model, with a coefficient of determination (R2) of 0.703. Virtual screening, using this model, identified five predicted potent inhibitors from the ZINC database comprising more than 300,000 molecules. The virtual screening recovered a known inhibitor, RN1, as well as four compounds where activity against LSD1 had not previously been suggested. Thus, we performed a machine-learning-enabled virtual screening of LSD1 inhibitors using only the structural information of the molecules.     

Application of self-organizing maps in compounds pattern recognition and combinatorial library design

In the computer-aided drug design, in order to find some new leads from a large library of compounds, the pattern recognition study of the diversity and similarity assessment of the chemical compounds is required; meanwhile in the combinatorial library design, more attention is given to design target focusing library along with diversity and drug-likeness criteria. This review presents the current state-of-art applications of Kohonen self-organizing maps (SOM) for studying the compounds pattern recognition, comparing the property of molecular surfaces, distinguishing drug-like and nondrug-like molecules, splitting a dataset into the proper training and test sets before constructing a QSAR (Quantitative Structural-Activity Relationship) model, and also for the combinatorial libraries comparison and the combinatorial library design. The Kohonen self-organizing map will continue to play an important role in drug discovery and library design.     

Cheminformatic Profiling and Hit Prioritization of Natural Products with Activities against Methicillin-Resistant Staphylococcus aureus (MRSA)

Several natural products (NPs) have displayed varying in vitro activities against methicillin-resistant Staphylococcus aureus (MRSA). However, few of these compounds have not been developed into potential antimicrobial drug candidates. This may be due to the high cost and tedious and time-consuming process of conducting the necessary preclinical tests on these compounds. In this study, cheminformatic profiling was performed on 111 anti-MRSA NPs (AMNPs), using a few orally administered conventional drugs for MRSA (CDs) as reference, to identify compounds with prospects to become drug candidates. This was followed by prioritizing these hits and identifying the liabilities among the AMNPs for possible optimization. Cheminformatic profiling revealed that most of the AMNPs were within the required drug-like region of the investigated properties. For example, more than 76% of the AMNPs showed compliance with the Lipinski, Veber, and Egan predictive rules for oral absorption and permeability. About 34% of the AMNPs showed the prospect to penetrate the blood–brain barrier (BBB), an advantage over the CDs, which are generally non-permeant of BBB. The analysis of toxicity revealed that 59% of the AMNPs might have negligible or no toxicity risks. Structure–activity relationship (SAR) analysis revealed chemical groups that may be determinants of the reported bioactivity of the compounds. A hit prioritization strategy using a novel “desirability scoring function” was able to identify AMNPs with the desired drug-likeness. Hit optimization strategies implemented on AMNPs with poor desirability scores led to the design of two compounds with improved desirability scores.     

Cheminformatics analysis of organic substituents: identification of the most common substituents,calculation of substituent properties,and automatic identification of drug-like bioisosteric groups

A large set of more than 3 million molecules was processed to find all the organic substituents contained in the set and to identify the most common ones. During the analysis, 849 574 unique substituents were found. Extrapolated to the number of known organic molecules, this result suggests that about 3.1 million substituents are known. Based on these findings the size of virtual organic chemistry space accessible using currently known synthetic methods is estimated to be between 10(20) and 10(24) molecules. The extracted substituents were characterized by calculated electronic, hydrophobic, steric, and hydrogen bonding properties as well as by the drug-likeness index. Various possible applications of such a large database of drug-like substituents characterized by calculated properties are discussed and illustrated by reference to a Web-based tool for automatic identification of bioisosteric groups.     

Drug-likeness scoring based on unsupervised learning

Drug-likeness prediction is important for the virtual screening of drug candidates. It is challenging because the drug-likeness is presumably associated with the whole set of necessary properties to pass through clinical trials, and thus no definite data for regression is available. Recently, binary classification models based on graph neural networks have been proposed but with strong dependency of their performances on the choice of the negative set for training. Here we propose a novel unsupervised learning model that requires only known drugs for training. We adopted a language model based on a recurrent neural network for unsupervised learning. It showed relatively consistent performance across different datasets, unlike such classification models. In addition, the unsupervised learning model provides drug-likeness scores that well separate distributions with increasing mean values in the order of datasets composed of molecules at a later step in a drug development process, whereas the classification model predicted a polarized distribution with two extreme values for all datasets presumably due to the overconfident prediction for unseen data. Thus, this new concept offers a pragmatic tool for drug-likeness scoring and further can be applied to other biochemical applications.

A new quantification method of drug-likeness based on unsupervised learning. The method only uses drug molecules as training set without any non-drug-like molecules.     

Rational design of inhibitors against LpxA protein of Acinetobacter baumannii using a virtual screening method

BackgroundAcinetobacter baumannii is a highly antimicrobial resistant nosocomial pathogen. Resistance to currently used antibiotics has limited effective drugs against this bacterium. This study aimed to propose a rational inhibitor design against the LpxA protein of A. baumannii using a virtual screening method based on a similar structure of ligands.MethodsIn this study, we targeted LpxA protein, which is involved in the early stage of LPS biosynthesis. In the next step, we used Peptide920 and 1,2- Ethanediol as templates to find similar compounds using Drugbank and Zinc15 webservers, respectively. Subsequently, molecular dynamics (MD) simulations were carried out for LpxA protein and two complexes of ZINC895081 and Macrolactam-1 which represented the highest binding affinity and best conformation. Finally, ADMET properties, water solubility and drug-likeness of the desired compounds were evaluated using SwissADME and DruLiTo softwares.ResultsAccording to considered criteria, Drugbank suggested 5 compunds including Ilomastat, Macrolactam-1, Macrolactam-2, Macimorelin, and Oglufanide. On the other hand, Zinc15 webserver suggested 4 compunds including ZINC895048, ZINC895081, ZINC901061 and ZINC1531008. The result of the HDOCK server and Molegro virtual docker (MVD) showed that Macrolactam-1 and ZINC895081 (Citrate) had the highest docking score. In addition, MD simulations showed that ZINC895081 and Macrolactam-1 ligands have the stable binding to the LpxA protein. According to Lipinski's rule, these two compounds are non-carcinogenic, non-toxic and promising inhibitors against LpxA of A. baumannii.ConclusionIt seems that Macrolactam-1 and ZINC895081 (Citrate) are two valuable promising inhibitors against the LpxA protein of A. baumannii. Further in vitro and in vivo experiments are needed to confirm the capabilities of these proposed compounds against A. baumannii.     

Identification of novel nt-MGAM inhibitors for potential treatment of type 2 diabetes: Virtual screening,atom based 3D-QSAR model,docking analysis and ADME study

In this study, a virtual screening procedure was applied to identify new potential nt-MGAM inhibitors as a possible medication for type 2 diabetes. To this aim, a series of salacinol analogues were first investigated by docking analysis for their binding to the X-ray structure of the biological target nt-MGAM. Key interactions for ligand binding into the receptor active site were identified which shared common features to those found for other known inhibitors, which strengthen the results of this study. 3D QSAR model was then built and showed to be statistically significant and with a good predictive power for the training (R² = 0.99, SD = 0.17, F = 555.3 and N = 27) and test set (Q² = 0.81, Pearson(r) = 0.92, RMSE = 0.52, N = 08). The model was then used to virtually screen the ZINC database with the aim of identifying novel chemical scaffolds as potential nt-MGAM inhibitors. Further, in silico predicted ADME properties were investigated for the most promising molecules. The outcome of this investigation sheds light on the molecular characteristics of the binding of salacinol analogues to nt-MGAM enzyme and identifies new possible inhibitors which have the potential to be developed into drugs, thus significantly contributing to the design and optimization of therapeutic strategies against type 2 diabetes.     

Structure-based de novo drug design using 3D deep generative models

Deep generative models are attracting much attention in the field of de novo molecule design. Compared to traditional methods, deep generative models can be trained in a fully data-driven way with little requirement for expert knowledge. Although many models have been developed to generate 1D and 2D molecular structures, 3D molecule generation is less explored, and the direct design of drug-like molecules inside target binding sites remains challenging. In this work, we introduce DeepLigBuilder, a novel deep learning-based method for de novo drug design that generates 3D molecular structures in the binding sites of target proteins. We first developed Ligand Neural Network (L-Net), a novel graph generative model for the end-to-end design of chemically and conformationally valid 3D molecules with high drug-likeness. Then, we combined L-Net with Monte Carlo tree search to perform structure-based de novo drug design tasks. In the case study of inhibitor design for the main protease of SARS-CoV-2, DeepLigBuilder suggested a list of drug-like compounds with novel chemical structures, high predicted affinity, and similar binding features to those of known inhibitors. The current version of L-Net was trained on drug-like compounds from ChEMBL, which could be easily extended to other molecular datasets with desired properties based on users'' demands and applied in functional molecule generation. Merging deep generative models with atomic-level interaction evaluation, DeepLigBuilder provides a state-of-the-art model for structure-based de novo drug design and lead optimization.

DeepLigBuilder, a novel deep generative model for structure-based de novo drug design, directly generates 3D structures of drug-like compounds in the target binding site.