Computational Methods for COVID-19 Treatment

Zahid Manzoor^{1, *}, Shaukat Hussain Munawar¹, Muhammad Farrukh Nisar², Muhammad Yasir Waqas², Muhammad Mazhar Munir¹

¹ Department of Pharmacology & Toxicology, Faculty of Bio-Sciences, Cholistan University of Veterinary & Animal Sciences, Bahawalpur, Pakistan

² Department of Physiology & Biochemistry, Faculty of Bio-Sciences, Cholistan University of Veterinary & Animal Sciences, Bahawalpur, Pakistan

Abstract

Computational approaches efficiently design the drugs to prevent diseases for which no drug is available. These techniques are also used for the development of new drugs. It involves using a variety of computer software for drug modeling and simulation, hence usually known as computer-aided drug designing (CADD). The computational tools provide crucial drug designing in a short period. These techniques are time and cost-effective as compared to conventional drug development methods. Computational methods can effectively model the suitable drug candidate by optimizing ligand-target interactions and observing the deep insight of cellular processes by its powerful tools. Several studies have applied these modern computational techniques to find out the possible therapy against the pandemic disease of COVID-19. The critical proteins of COVID-19, including 3C-like protease, papain-like protease, and RNA polymerase, are targeted to model the effective drug. CADD approaches suggest anti-viral drugs, anti-coagulant, anti-HIV drugs, and anti-fungal drugs to have little effect against COVID-19. This chapter aims to overview the different CADD approaches to design the possible drug for the treatment of COVID-19.

Keywords: CADD, Computational methods, COVID-19, Treatment.

^* Corresponding author Zahid Manzoor: Department of Pharmacology & Toxicology, Faculty of Bio-sciences, Cholistan University of Veterinary & Animal Sciences, Bahawalpur, Pakistan; E-mail: zahidmanzoor@cuvas.edu.pk

INTRODUCTION

The development of a novel drug usually takes a long time and is often associated with a high risk of failure and high cost. Typically, the complete drug development procedure, from the drug concept to the market, may take about 14 years [1]. The drug development procedure is not an easy task, and the whole process may involve a very high cost ranges from 0.8 to 1.0 billion USD [2]. However, advanced scientific approaches expedite this time-consuming procedure

in a short time [3, 4]. In the last few years, heavy investment has been made in the drug discovery procedure, but the output was not up to the expectations because of the low efficiency and high failure rate in the process [3].

Consequently, scientists and researchers searched for alternate tools to increase the success rate with high accuracy in a short research period. Various scientific approaches have been developed for this purpose. Among these techniques, computer-aided drug design (CADD) is one of the most effective methods for drug discovery. CADD is a widely used term that refers to the different computational tools for designing the compounds. It usually includes the analysis, storage, management, and modeling of compounds. CADD covers almost all of the essential aspects of the drug discovery and development process. Different computer-integrated programs are used to design compounds, assess the most appropriate candidate, and develop digital repositories for studying chemical interactions [4]. The application of CADD has altered the pipeline for drug discovery and development. CADD tools potentially identify the drug targets, validate and optimize the procedures, and may be applied in preclinical studies [5, 6].

Fig. (1))

The major steps involved in the CADD.

CADD approach is cost-effective and could reduce costs by half than conventional drug discovery and development [7]. The standard approaches of CADD can be divided into structure-based drug design and ligand-based drug design techniques.

The important steps involved in the CADD approach are shown in Fig. (1) . The CADD technique is generally divided into two important categories. The first category is the structure-based drug design, while the other refers to the ligand-based drug design. Structure-based design is usually applied to the availability of the 3D structure of a molecule. In this technique, the de novo approach provides valuable information. If the 3D structure of the target is unknown, then a ligand-based drug design can be applied. In this approach, pharmacophore modeling and quantitative structure-activity relationship (QSAR) can provide a deep insight into ligand-target interactions, which may be further processed by Virtual screening, selection of suitable compound, lead discovery, and ultimately the modeling of the new drug [8]. The purpose of this chapter is to highlight the importance of CADD approaches and the possible use of these tools in exploring the novel treatment for COVID-19.

Integrated Pipelines of COVID-19

Computational methods are more convenient in terms of less time-consuming and more cost-effective. These modern techniques have advantages over the historical practices of drug development in the pharmaceutical industry. In conventional methods, a series of experiments must be performed to find the suitable compound with desired properties. These experiments are usually time-consuming. With the advancement of technology, CADD tools efficiently analyze several compounds and quickly sort out the compound of interest to be evaluated for further biochemical reactions. Modern technology enables us to search for the drug target by the visual image of the 3D structure of the compounds.

In the current scenario of COVID-19, it is crucial to find an effective and safe drug very quickly; however, the potential drug candidate for the treatment of COVID-19 could be identified by using the modern computational screening of drug libraries. Virtual screening (VS) of ligand databases plays a vital role in exploring suitable molecules and has accelerated the initial stages of drug discovery [9, 10]. The purpose of vs is to rapidly recognize the potential hit molecules and can be tested experimentally and clinically. The crystal structure of molecules provides essential information in the identification process of potential drug molecules. vs is used by docking drug fragment libraries where the crystal structure of molecules is available. If the crystal structure is not available, then homology models are used. vs is used to identify the inhibitors in this scenario [11].

Virtual Screening of COVID-19

Two important proteases are coded by SARS-CoV-2 polyprotein. These proteases play an essential role in the process of non-structural protein (NSPs) and their release. The main protease is 3C like protease, also known as M_pro, whereas the other is a papain-like protease called Pl_pro [12]. Both proteases are considered to play a vital role in SARS-CoV-2 pathogenesis and hence an important target for drugs against SARS-CoV-2 and other coronaviruses [13, 14]. The crystal structure of only M_pro of SARS-CoV-2 is available and published in Feb 2020 (PDB ID 6lu7). The leading virtual screening against COVID-19 was performed in this study. The findings of this study will provide useful information to apply further CADD approaches to design possible effective drugs against COVID-19.

Another study employing virtual screening and targeting the crystal structure of M_pro of SARS-CoV-2 was conducted. Many drugs that the FDA approved were screened in this study. The study also provides important information about SARS-CoV-2 by conducting sequence statistics and phylogenetics. The results showed that SARS-CoV-2 M_pro formed a distinct phylogenetic group with SARS COV. It was different from MERS-COV. Sequence comparison analysis revealed that SARS-CoV-2 resembles about 96.06% with SARS and about 51% with MERS-COV. Some important drugs that were virtually screened against COVID-19 include telbivudine (used against the hepatitis-B virus), ribavirin (anti-viral drug), nicotinamide (vitamin), and vitamin B₁₂, etc . Data from the study revealed that one of these combinations of these essential drugs could provide better protection against COVID-19 [15].

Drug Libraries for COVID-19

To combat the SARS-CoV-2 infection, some drugs that inhibit the HIV protease such as Lopinavir, are suggested. Though their therapeutic potential against coronaviruses has not been determined up till now. In SARS-CoV-2, the protease protein is an important target to inhibit the pathogenesis of this pandemic disease. In the current study, CADD techniques are used to model an appropriate drug by screening the Marine Natural Product (MNP) library. Two important CADD techniques, molecular docking and the hyphenated pharmacophore model, were used to screen the MNP library. Molecular dynamics and re-docking techniques further confirmed the results of the study. The study proposed the 17 important protease inhibitors originated from natural marine origin. The crucial CADD technique identified these substances. Hence, it is suggested that one of these substances could provide better protection against COVID-19 [16].

Drug Repurposing Approach for COVID-19

Drug repurposing or repositioning is a promising approach to find out the possible ways to improve the therapeutic potential of the existing drugs. It may involve the failed developing compounds for further clinical trials. This approach is expanding rapidly for the treatment of rare and neglected diseases. Drug repurposing mainly focuses on drug-drug interaction (DDI) and drug-drug target interaction (DTI). One of the surveys on DTIs collected by the Drug Bank databases showed that a drug has many drug targets. A drug can bind to three different sites present on the other drug molecule [17]. The study concludes that polypharmacology is a common phenomenon of most drugs. Hence, it is crucial to find out the potential DTIs for both drug candidates and approved drugs. The basis of drug repurposing is to identify the DTIs. If a drug is selected without knowing the DTIs, it may cause specific side effects. Polypharmacology provides a new approach to drug designing with the least side effects and is more useful therapeutically. CADD has been playing a vital role in modern drug designing. A hieratical strategy consisting of various types of scoring functions has been implemented during drug lead identification and optimization phases to enhance the computational efficiency and accuracy of CADD. In addition, certain docking scoring functions, such as one employed by the Glide docking program, are very efficient for screening an extensive drug library [18]. However, it is not so much accurate in functioning. Another scoring function is the molecular mechanical force field (MMFF)-based scoring functions. It is more accurate, but the efficiency of MMFF is poor.

A hierarchical virtual screening (HVS) has recently been developed to improve the efficiency, accuracy, and success rate of rational drug designing [19, 20]. The crystal structure of SARS-CoV-2 revealed the target protein where different drugs might interact to decrease the pathogenicity of SARS-CoV-2 infection. Multiscale modeling techniques such as Flexible docking and MM-PBSA-WSAS were applied as filters to identify the potential drugs as inhibitors of COVID-19. CADD-based approaches are more efficient than experimental techniques in exploring the possible treatment for epidemic disease outbreaks like COVID-19.

Drug repurposing is an effective tool of CADD that can shorten the time and cost for the discovery of the new drug. This technique modifies the existing drug to model the structure of the new drug. An experimental study was performed on the available anti-viral drugs by applying the drug repurposing approach. The drug target, protein-protein interaction (PPI), and structural sequence were also evaluated. The results of the phylogenetic analysis showed that the SARS-CoV-2 genome was about 79% similar to the SARS-COV. Notably, the sequence of viral envelops and nucleocapsid of SARS-CoV-2 and SARS-COV were very similar and shared about 96% and 89% similarities. In this study, 16 different drugs were repurposed, including sirolimus, mercaptopurine, and melatonin against SARS-CoV-2, further validated by different CADD techniques. Data from the study proposed three different drug combinations emodin with toremifene, melatonin plus mercaptopurine, and dactinomycin with sirolimus that can show effective protection against COVID-19 [21].

Molecular Databases of COVID-19

In a study, in silico approach was used to predict the potential drugs against COVID-19 using molecular databases such as the ZINC database and CHEMBL database. The co-crystallized structures of 6W63 and 6Y2F from the protein data bank (PDB) were studied. Approximately 700 various compounds from CHEMBL/ZINC databases and 1400 substances from PDB were chosen based upon their possible interactions with the documented binding site. Various computational approaches were applied to screen about 300 potential candidates from different databases. Also, approximately 66 FDA-approved drugs were screened by employing these techniques. Some of the essential drugs include darunavir, lopinavir, ritonavir, and cobicistat, etc . These drugs and compounds may provide adequate control against COVID-19 after further CADD analysis [22].

Genomic Information of SARS-CoV-2 (COVID-19)

The genomic studies of SARS-CoV-2 revealed that the 3-chymotrypsin-like cysteine protease (3CL_pro) enzyme is vital in controlling the viral replication, and the virus requires this enzyme to complete its life cycle. It was proposed in the recent studies that 3CL_pro could be the right target for new drug development against MERS-COV and SARS-COV infections. In recent studies, it was suggested that the genome sequence of SARS-CoV-2 is parallel to the genomic sequence of SARS-COV. Hence, the sequence of 3CL_pro was analyzed, and the 3D model was constructed using the homology model. The constructed model was screened against a huge Data Bank of the medicinal plant containing about 32,297 potential anti-viral and bio-active phytocompounds. It was proposed in the study that the top nine screened compounds may provide useful information for the drug development process against COVID-19 [23].

Homology Modeling and Molecular Docking of COVID-19

Homology modeling refers to the comparative modeling of protein. It involves identifying one or more protein structures that show a similar structure to the query sequence. The residues are aligned in the query sequence to make the template sequence-structure [16]. Proteins with similar structures are evolved to have a similar sequence as naturally occurred homologous protein and thus are more conserved to generate a structural model of the target with the help of template structure and sequence alignment techniques.

Molecular Docking

Molecular docking is a computational method that predicts the best orientation of a molecule to the other molecule during the formation of a complex. The other molecule may be a protein or a ligand. Docking refers to the ligand-binding either to its target protein or to its receptor. This technique is mainly used to predict the interaction of stable drug molecules by observing and modeling the interactions of drug molecules between drugs or their target proteins. Many possible ligand conformations and orientations are developed during the molecular docking process, and the most suitable ones are selected for further research [24].

There are numerous tools for molecular docking; some of the common tools include FRED, AutoDOCK, FTDock, and eHITS. Docking studies are helpful in a deep understanding of the type of interaction between the different drug molecules and proteins. Novel algorithms are available at present, which provide a deep insight into the microenvironment during an interaction. These tools also predict the presence or absence of water molecules during the exchange. The available computational tools have remarkably increased the flexibility of computation possibilities. Molecular docking has enabled us to study the small molecule docking such as protein-protein, protein-RAN, and protein-DNA interactions considering the associated cellular complexities [25, 26]. Studying complex interaction using a distributed computation is a common practice in computational biology. Recently, homology modeling and in silico docking are applied to some possible medications to determine the potential treatment of the pandemic outbreak of COVID-19.

The structure of SARS-CoV-2 closely resembles SARS-COV. The advantage of these known proteins is to build a proposed model for the treatment of COVID-19 [27] quickly. The drug development by the conventional method may take many years for COVID-19; hence the CADD approach was used to develop the model drug. The two important proteins of SARS-CoV-2 were targeted for in silico docking model. These include spike glycoprotein and the SARS-CoV-2 3CL main protease.

The spike protein of SARS-CoV-2 interacts with the receptor protein of the host cells and enters into the cells. The CADD tools revealed that the entrance mechanism of the SARS-CoV-2 spike protein is similar to the SARS virus. The COV spike protein interacts with the host cell receptor, which allows the spike protein to enter into the host cell. The interaction of the COV spike protein to the host receptor is also similar to that of the SARS virus. The ligand-protein has a great affinity to the angiotensin-converting enzyme2 (ACE2) [28]. The modern computational approaches revealed that the COV spike protein is structurally similar to the SARS spike protein. The receptor interaction of both proteins is also very similar. However, a 27% variation exists between the interaction mechanisms of both proteins. These variations are mostly from the host cell. There is no crystal structure COV spike protein available until now; hence, homology modeling is used to predict COV spike protein crystal structure using SARS spike protein (PDB; 2GHV) as a model template.

The second important protein of SARS-CoV-2 is the 3CL_pro main protease (PDB: 6LU7) [29] associated with many controlling mechanisms of the virus. This protein is highly conserved in SARS-CoV-2 as compared to the SARS virus [30]. An essential function of 3CL_pro protease is linked with the replication of viruses which may be an ideal target for the action of new drugs [31].

Spike proteins and protease equally contribute to viral transmission and pathogenesis. To decrease the virulence of SARS-CoV-2, both of these proteins or anyone protein can be targeted. Some efforts have been made in this regard. A large data of bioactive compounds with known structures were run to attach different binding sites of COV spike protein and catalytic sites of protease protein using computational techniques. One of the possible solutions to combat SARS-CoV-2 virulence is to conduct quick trials with approved compounds by the FDA with modern computational approaches. CADD tools revealed that saquinavir, indinavir, remdesivir, and zanamivir are potential candidates to target 3CL_pro main protease of SARS-CoV-2. It is important to note that adeflavin, flavin adenine dinucleotide (FAD), coenzyme-A, and B₂ deficiency medicine may protect against COVID-19.

Protein Modeling

Protein modeling is an important tool in CADD to predict the 3D structure of most of the proteins from their amino-acid sequence. Various techniques such as NMR spectroscopy and X-ray crystallography have been used worldwide to determine the protein structures in detail. Characterization of the amino-acid sequence in protein structure is one of the challenging tasks. This difficulty is usually addressed by the careful observation of the 3D structure of a protein. Two important principles of science: the theory of evolution and the laws of Physics have provided much guidance in studying the 3D structure of natural proteins. The application of both these principles gave rise to different prediction approaches for a protein structure [32].

Many protein conformations are made to assess the lifespan of protein. The first method mostly relies on the amino-acid sequence and is called ab initio or de novo method. This type of prediction mainly focuses on the sequence pattern and does not depend upon the similarity of folded protein between the modeled protein and any other known structures [33]. The de novo approach assumes that the native biomolecule structure corresponds to the free energy; hence, it is challenging to predict the life span of the protein. There are two main components of the de novo method. One is the efficient procedure to carry out the conformational search, and the other is the free energy function which is used for the evaluation purpose of conformational changes.

The second method is mostly associated with the detectable similarities between the model sequences with any known structure. Comparative modeling and threading are some examples of this method [34, 35]. The structure of one protein in a family is determined experimentally. It will provide a guide to model the other protein members of the family. The primary alignment structure plays a vital role in this phenomenon. In the process of protein modeling, a 3D structure of unknown protein usually target protein is built depending on one or more related known structures of protein which are referred to as templates [36, 37]. For successful modeling, the detectable similarity between the template and target sequence and availability of correct alignment is crucial. The comparative approach can model protein structure because a slight change in the amino-acid sequence may change its 3D structure, which is easily detectable [38]. Protein structure can be modeled because the protein in a family is more conserved than their prime sequences [39]. Thus, if the sequence similarity of two proteins is detected, their structural similarity is likely to be assumed. It should also be noted that some proteins show similarities many times. But actually, they have very little structure similar or an equal portion is non-detectable. In this scenario, the proteins are assumed to be identical to each other.

Each of these methods has some pros and cons. But if we compare both of these methods, then comparative modeling seems to predict the 3D structure of an unknown protein more reliable than ab initio protein structure prediction [33, 35].

The following are some important steps in protein modeling:

· To explore the most suitable proteins with known 3D structures that have a similar sequence to the target protein

· Selection of the template structures

· Alignment of the template sequences with the target protein sequence

· Based upon the template structure, build a model that best fit the target sequence

· Evaluate and validate the proposed model.

These important steps can be repeated again and again until a desired perfect model is achieved. SARS-CoV-2 has an important transmembrane spike (s) glycoprotein that facilitates the coronavirus entry into the host cells. Spike glycoprotein produces some sets of a glycoprotein called “homotrimers” that protrude out of the viral surface [40]. There are two sub-units of S-glycoprotein named S₁ and S₂. These sub-units are responsible for the attachment of viruses to host cells. After attachment, the viral membrane is fused with the host cell membrane. Different coronaviruses use different domains within the S₁ sub-unit for the entrance to the host cells. These domains are named S_A and S_B. SARS-CoV-2 and SARS-related coronaviruses enter the host cells via the interaction ACE2 by the domain S_B [41, 42]. Murine polyclonal antibodies may inhibit the entry of SARS-CoV-2 mediated by SB. It is expedited from the data that the vaccine having neutralizing antibodies could be a possible tool to combat coronavirus infections, including COVID-19 [43].

It was evident from the previous studies that a positive selective pressure builds on the spike glycoprotein, nucleocapsid, and ORF1ab, but there is no such evidence in SARS-CoV-2 until now. However, recent studies showed that the presence of pervasive episodic selection in 2 sites. The presence of Gln residue was found instead of Asn at the amino acid position of 380 of the Wuhan coronavirus sequence. Similarly, a Thr residue was found instead of Ala in the amino acid position of 140. These changes create negative pressure in 6 sites (14%) confirmed by FUBAR (Fast Unconstrained Bayesian Approximation) analysis.

Similarly, a remarkable pervasive episodic selection was found in the spike glycoprotein region because of the two different sites at 536 and 644 nucleotide positions compared to the reference sequence. There was Asn residue present instead of Asp acid at 536 positions of COVID-19, while a Thr residue was present at 644 positions instead of an Ala residue. A remarkable pervasive negative selection was confirmed by FUBAR analysis in 1065 sites that appropriately contributes to 8.7% [44].

There is some evidence for the positive selective pressure in the glycoprotein of SARS-CoV-2. Data have shown that an Asn residue was present in amino acid position 536 in SARS-CoV-2, while a Gln residue was present in Bat SARS-like coronavirus and Asp residue was found in SARS virus. The difference also exists at amino acid position 644. A Thr residue was present in SARS-CoV-2, while a Ser residue was present in Bat SARS-like coronavirus, and Ala residue was present in the SARS virus. Another study showed that there is variation in some key residues linked with the binding mechanism to the ACE2 receptor in SARS-CoV-2 receptor-binding domain. These include Asn501, Asn439, Gly485, Phe486 and Gln493, and SARS-CoV-2 numbering). Some missing residues in amino acid positions 455–457, 463–464, and 485–497 occurred in the bat-derived strains [45].

Some of the sites are also found that create positive selective pressure in the ORF1ab region. Significantly, amino acid position 501 is important in this regard. A Gln residue is present in SARS-CoV-2, while Thr residue is present in bat SARS-like coronavirus, whereas an Ala residue is present in the SARS-CoV-2 virus. The SARS-CoV-2 has a Ser residue at the amino acid position of 723. The Gly residue is present in the other two viruses of bat SARS-like virus and SARS virus. The SARS-CoV-2 has a pro residue at position 1010 of the amino acid sequence. A His residue is present in the bat SARS-like coronavirus, and an Ile residue is present in the SARS virus.

Regarding the residue position in 723, especially at 543 in the nsp3 protein, SARS-CoV-2 exhibited a Ser residue instead of Gly present in SARS coronavirus and the Bat SARS-like virus. In this scenario, it can be suggested that this replacement may enhance the firmness of the polypeptide chain in terms of a steric effect and the ability of the Ser side chain to make hydrogen bonds. Moreover, Serine residue has a special effect. It may act as a nucleophile in the known structures, or it might be active site for some enzymes, and it can be a site for the phosphorylation process. But one of the CADD approaches suggests that it is complicated for most of the solvent to identify this position.

Suppose we look into the detail of amino-acid position 1010 that contributes to 192 of the nsp3 protein. The common region of SARS virus and bat SARS-like coronavirus has a non-polar amino-acid a polar amino-acid, whereas a pro residue is present in SARS-CoV-2 at this position. This scenario expedites that the stiffness and steric bulge of pro leads to a change in the molecular shape of SARS-CoV-2 compared to the other two viruses.

In SARS-CoV-2, a mutation was observed near the polyprotein domain in nsp3, parallel to the presence of phosphatase in the SARS coronavirus (PDB code 2ACF). These are considered to have an important role in viral replication in the host cells [46]. Some of the CADD approaches like I-TASSER suggest that some of the solvents could partially reach these positions. Moreover, positive selective pressure in this region relates to some of the clinical importance of SARS-CoV-2 compared to bat SARS-like coronavirus and SARS virus. It is necessary to find the exact location of amino acid sequence change in SARS-CoV-2 by using modern CADD tools to predict the actual protein portion involved in viral entry and pathogenesis. It would be useful in the deep understanding of the SARS-CoV-2 action and its virulence.

Furthermore, it is essential to identify the site of exact molecular evolution in SARS-CoV-2, which would be helpful in the drug development or preparation of a vaccine against this pandemic disease. One possible reason for the more contagious behavior of SARS-CoV-2 could be the structural similarity of the region where positive selective pressure exists, and the stability of the mutation occurred in the endosome-like protein domain of the nsp2 protein as compared to other coronaviruses. The other important destabilizing mutation that happened near the phosphatase domain of the nsp3 protein may provide a reason for slower viral replication of SARS-CoV-2 than the SARS virus. However, further detailed studies are required in this regard [47].

Fold and Function Assignment System

The nsp3 protein of SARS-CoV-2 has many functions and contains different domains. Many important functions such as interference with host immune response and cleavage of the viral polyprotein, replication, and transcription of viral genomic material are assisted by nsp3. The fold and function assignment system of bat coronavirus HKU9 revealed that a unique region called the “SARS-unique region” is present in the NMR structure of the protein. The detailed study of protein showed that a double-wing motif or frataxin fold (a1b fold) was present, which was supposed to bind with DNA or metal ions and interact with other proteins. The structure of the protein was very similar to the SARS coronavirus nsp3 protein. Moreover, biochemical and bioinformatics analyses identified a functional site that was conserved in a few betacoronaviruses. The presence of an active site supports the statement that the “SARS-unique” region is conserved only in some phylogenetic groups of coronaviruses and not related to the human SARS virus [48].

Protein Data Bank

In a study, some nucleoside analogs and HIV protease inhibitors were repurposed by the molecular docking approach to design the possible drug to treat COVID-19. RosettaCommons and AutoDock Vina evaluated docking scoring. The results showed that Remdesivir and Indinavir have the best docking scores and perfectly fit in the protein pockets. Contrary to this, the active form of remdesivir showed a perfect dock on the functional sites of the NTP binding motif. The detailed investigation revealed that Indinavir does not dock on the functional sites of protease; hence the efficacy of Indinavir is not good against COVID-19. However, further detailed clinical studies are required before issuing the final recommendation [49].

Pharmacophore Modeling

Pharmacophore modeling is an essential tool for CADD. It is a very versatile technique and can be used in various situations. This approach is mostly used for the rational drug designing of novel drug compounds. Pharmacophore was internationally accepted by the International Union of Pure and Applied Chemistry in 1977. It may be defined as “It is the ensemble of electronic and steric characteristics which are essential to secure the optimal supramolecular interactions to a specified biological target and to enhance or stop the biological action” [50].

The pharmacophore is considered one of the most common and largest characteristics of the molecular interaction of biological substances. Pharmacophore is a concept; it does not stand for an actual biological substance for interactions. Even though the definition of pharmacophore is very clear, some people have used this term to represent a drug molecule such as sulfonamides or prostaglandins or bioactive compounds such as flavones in medicinal chemistry [51].

In the early stage, the concept of pharmacophore was limited to some computer programs. Over time, it has become an important tool for CADD. Pharmacophore is related to any molecular recognition exhibited by any atom or any substance in a molecule. These molecular interactions may involve hydrogen bond acceptors or donors, anionic, cationic, hydrophobic or aromatic, or any other possible interactions [51].

Various molecules under investigation may be tested at the pharmacophore level. This practice is known as “pharmacophore fingerprints”. Sometimes pharmacophore testing/fingerprints are limited to few characteristics in a 3D model. In this scenario, pharmacophore is referred to as a “query”.

Pharmacophore Fingerprinting

Pharmacophore has the option to limit its functions to test the 3D or 2D molecule. If any 3D molecule undergoes pharmacophore fingerprinting, the software has the opportunity to change the setting [52, 53]. Pharmacophore fingerprinting is a unique way of testing that interprets a molecule in special data sets. In the process of pharmacophore fingerprinting, all of the possible available points are taken into consideration while testing a molecule [54]. In topological fingerprints, the distance between the various molecular features is measured and denoted as bonds. The result of the pharmacophore fingerprint is in the form of a string that shows various possible positions where the molecule may interact with other substances.

There are several different pharmacophore fingerprints available that are used according to the different situations. Some fingerprint is designed in a way that it can be used to predict the various molecular similarities among different molecules. Pharmacophore fingerprinting may also determine the similar features of active ligands that may help estimate the key contributing features of biological action.

Pharmacophore Model or Query

Some of the pharmacophore models are specially designed in a 3D pattern [55]. There are different ways to label various features. It may be a single feature or a combination of various similar features usually displayed as “AND”, “NOT”, and “OR” to merge multiple patterns of interaction in one label. There are some additional features usually used to describe the boundary of a receptor. These features are essential to screen a library of small molecules [56]. All the molecules show their low level of energy conformations in these libraries.

These conformations are used to create the pharmacophore query by adjusting the pharmacophore and relevant compound. The pharmacophore query consists of many spheres. If a molecule exactly fits in the sphere and fulfills the query features, it is known as a hit molecule. The pharmacophore query is not so simple it's a complex procedure. It is tough to find the hit molecule in the given libraries. Hence, partial matching may be used to decrease the complexity of the pharmacophore query. Only a few features that are most relevant to an activity are matched in such cases. These models are further aligned with the other tools of CADD, such as molecular docking simulations [57, 58].

Modeling Softwares

Some of the important modeling software used in the CADD approach are given below in Table 1.

Table 1 Different modeling software in the CADD.

S.No.	Name of Modeling Software	Main Features	Uses
1.	MapCheck	To compare the dose or measure the fluency	Calculate the Pharmacokinetic parameters
2.	DDD Plus	To study the disintegration and Dissolution properties	Estimate the PK parameters
3.	GastroPlus	Estimate the In-vivo and in-vitro (IVIV) correlations of various preparations	Use in the pharmacokinetic studies
4.	Schrodinger	Ability to dock the Ligand-receptor binding	Molecular dynamics and interactions of ligand
5.	AutoDock	Estimate the interactions of the protein with ligand	Molecular dynamics and interactions of ligand
6.	BioSuite	Powerful analysis of genome sequence	Molecular dynamics and interactions of ligand
7.	GOLD	Docking study of the ligand with proteins.	Molecular dynamics and interactions of ligand
8.	ArgusLab	Calculate the molecular docking and build the models of molecules	Structure-activity relationship (SAR) and molecular modeling
9.	Maestro	Analysis of the molecules models	Molecular modeling
10.	SYBYL-X Suite	Build the models on molecules and design the models relevant to ligands	Molecular modeling
11.	GRAMM	Docking of protein to protein and protein to a ligand	Structure-activity relationship (SAR) and molecular modeling
12.	PASS	Built and analyze the SAR models	Structure-activity relationship (SAR)
13.	Sanjeevini	Estimate the binding affinity of protein with ligands	Molecular modeling
14.	Discovery Studio® Visualizer	Predict the protein structure	Visualization and analysis of images
15.	Xenogen Living Image Software	Present and analyze the in vivo images	Visualization and analysis of images
16.	AMIDE (A Medical Image Data Examiner)	Analyze the images in molecular models	Visualization and analysis of images
17.	REST 2009 Software	Analyze the data of gene expressions	Analysis of large data sets
18.	QSARPro	Study the interactions of different proteins	Analysis of large data sets
19.	GeneSpring	Powerful capacity to identify the minor variation within the sample and rectify it.	Analysis of large data sets
20.	MARS (Multimodal Animal Rotation System)	To study the activity of enzymes and animal response. To track the nanoparticles and their delivery mechanism.	To study the behavior
21.	Ethowatcher	Critically examine the response of animals	To study the behavior

Monte Carlo Method

The Monte Carlo simulation technique is advantageous and is being used in a variety of computational methods. It is based on large numerical numbers for calculations. In the CADD, this technique is commonly employed to test the Boltzmann distribution and build a Markov Chain Monte Carlo [59], particularly with the Metropolis algorithm [60]. It can also be used to explore the bound structure conformations.

Modeling the protein-ligand complex can be possible using the all-atoms force field paradigm [61], resulting in a very high dimensional phase. In this case, it is tough to estimate a parameter. The building of the Monte Carlo simulation overcomes this difficulty if the relative probability is known [62]. In this scenario, Monte Carlo provides an alternation approach to calculate the molecular dynamics [63]. The routine simulation takes a very high cost for the computational evaluation of the model, and it takes a very long time for completion. The Monte Carlo method has resolved this issue. The use of the algorithm appropriately evaluates the data according to the need. It enables us to use the Monte Carlo method more efficiently in drug designing. However, the main challenge in the Monte Carlo method is to generate uncorrelated poses having a significant weight which usually produces ruggedness of the energy landscape. This energy can be measured by the force field. The possible cause of this difficulty might be ligand or protein flexibility.

The Stochastic Global Optimization Procedure

In the last few decades, stochastic global optimization procedure has gained much attention in drug designing. This procedure deals with any kind of issue having a discrete variable, non-differentiated functions, or continuous variables during the processing. It has also been applied in biochemical and environmental processes that show different algorithms.

Pseudo-Brownian Positional/Torsional

The molecular docking method usually predicts the interaction of two independent protein structures, which typically fail on a large set of data. The possible reasons for failure can be the complex methods, including determination of scoring functions, difficulty in simulating the data, and rearrangement of residues at binding sites. The pseudo-Brownian positional/torsional approach minimizes these problems by observing the interaction of side-chains and conserving the energy.

Protein-protein Docking

The most important phenomena in the majority of cellular reactions are PPI. It has been estimated that about 4 million protein interactions occur in the human body [65]. PPI forms the basic foundation of most biological processes. It also includes two important processes signal transduction and regulation of proteins. These processes dominate over the other processes in increasing the complexity of cells [64]. These processes are fundamental. A mild error in these processes may result in immune diseases or increase the proliferation of cells [64].

For the computations studies of protein, the 3D structure of the protein is a prerequisite. The X-ray technology provides some insight into the protein, but this approach has few limitations. Hence, the best choice to determine the 3D picture of protein-protein complexes is the computational techniques [66, 67]. In these techniques, protein-protein docking (PPD) provides reliable data of the protein molecules.

Small molecule docking has been extensively used in modeling studies for a long time [68]. However, PPD is more complex than small molecular docking. Proteins are usually dynamic. It could be one of the possible reasons for the complexity of the PPD method. Typically, most of the proteins are not stable. They often fluctuate to a varying degree which increases the complexity for docking study the proteins. Determine the binding site in protein is also challenging.

PPD is still the most common method in docking studies despite these challenges [67]. There is an increasing trend in the publication of PPD studies. For docking of proteins, detailed information regarding the rotation and translation position of the binding molecules is required. This information can be collected by other CADD techniques [69].

It has been assumed that PPD can be performed by knowing the shape of the protein molecules and amino acid sequence [67]. Some other CADD techniques, such as Rigid-body docking, facilitate the PPD in locating proteins' fixed and transitory positions [70]. Typically, the PPI procedure involves the following important steps [71].

· To search for the appropriate stage when the protein-protein complex is made

· The sampling time involving the different scoring criteria of the protein complexes

· To refine the final solution based upon various scores.

The protein complex is scored according to its geometrical structure, Physico-chemical nature, hydrogen bonding interactions, and other important information needed for the docking procedure [70]. The majority of the PPD methods relied on the rigid-docking procedure in the past. Currently, the flexibility of the side-chain protein is also taken into consideration. This feature is very important in PPD experiments. The majority of experiments conducted in CAPRI failed due to the wrong measurement of protein conformational changes [67].

Drug-target Interaction

The identification of drug-target interaction (DTI) has a fundamental role in drug development. In general, drugs interact with one or more proteins as their target. However, discovering the novel drug interaction to their target is essential as the divergent behavior may result in some unwanted effects [72]. In recent years, scientists have tried to explore the DTI by conducting clinical experiments and observation. Conducting the clinical experiments is quite time-consuming, and the high attrition rate is the main hurdle in these experiments [73]. Hence, the CADD approaches to identify the DTI have gained much popularity in the last few decades.

Some large data about various DTIs have been discovered and placed in databases accessible to the public, offering them an excellent hands-on prediction of the drug-interaction model. Various databases used for DTIs have been developed and used for the public, such as Matador and SuperTarget, Therapeutic Target Database, Kyoto Encyclopedia of Genes, Genomes (KEGG), and DrugBank. These databases provide useful information for computational methods [74, 75].

Some of the CADD tools for DTIs are literature text mining methods, ligand-based methods, and docking simulation [76, 77]. But there are some restrictions/limitations for all of these three methods. The docking simulation method is successful only on a small scale because it needs a 3D profile of the target protein available only for a small portion of protein.

Recently, various commercially available anti-viral drugs are modeled through Molecule Transformer-Drug Target Interaction (MT-DTI) to predict the possible binding of these drugs to the viral protein of COVID-19. It was found that atazanavir, which is an antiretroviral drug used against HIV infection, showed the best inhibitory potency with Kd of 94.94 nM against 3C-like proteinase of COVID-19. The other drugs remdesivir, efavirenz, ritonavir, and dolutegravir, also showed good potential with Kd values of 113.13 nM, 199.17 nM, 204.05 nM, and 336.91 nM, respectively. Interestingly, some other drugs, such as darunavir, lopinavir, and ritonavir, are also designed to hit the COVID-19 protein. Data suggest that these drugs can bind with the COVID-19 replication complex with inhibitory potency of Kd below 1000 nM. Kaletra (lopinavir/ritonavir) was also found to have the potential to inhibit COVID-19. It was proposed to consider all anti-viral drugs selected by the MT-DTI model while developing therapies against COVID-19 [78].

Molecular Dynamic Simulation

In a study, 16 approved drugs by the FDA and large data sets of compounds consisting of about 1000 bioactive compounds from the Asinex Focused Covalent (AFCL) library were screened to target the 3CLpro protease SARS-CoV-2. The compounds showing remarkable better docking scores and stable interactions were selected. By applying the CADD approaches 621 potential compounds were identified having the ability to bind with SARS-CoV-2 3CLpro. A molecular dynamic (MD) simulation approach of 50 nanoseconds was applied to stabilize the binding between the SARS-CoV-2 3CLpro and identified bioactive compounds. These compounds may provide adequate control against COVID-19 after the conduction of further necessary studies [79].

The binding affinity of noscapine (23B)-molecular dynamic simulations examined protease of SARS-CoV-2 at various temperatures. Results showed that noscapine has an anti-viral effect. The first study explicit the anti-viral potential of noscapine compound based upon the molecular dynamic simulations [80]. In another study, the proposed anti-viral compound was investigated by molecular dynamic simulation and docking studies against COVID-19.

Binding Affinity

Recently, the binding mechanism of the novel SARS-CoV-2 spike protein to the host cell surface receptor (Glucose Regulated Protein 78 (GPR78) was predicted using the modern molecular modeling technique. The SARS spike protein was modeled in this experiment. CADD tools identified four regions that have similar sequences and functions to cyclic Pep42. The PPD approach was used to predict the best fit of the region in the GRP78 Substrate Binding Domain β (SBDβ). Data revealed that the receptor-binding domain and SBDβ of GRP78 have a critical role in identifying host cells' receptors. It is also speculated that SARS-CoV-2 was more likely to bind at region-IV (C480-C488) and in region III (C391-C525) and at the GRP78. Region IV was assumed to play the major factor which forces the GRP78 to bind with an estimated binding affinity of -9.8 kcal/mol. The predicted nine residues could provide important data to develop the specific therapy to combat COVID-19 [81].

Binding Free Energy Calculation

The stilbenoid analogs were repurposed by molecular docking and estimating the binding free energy and molecular dynamic simulation to search for the potential candidate for SARS-CoV-2 infection. Based upon the CADD approaches, four potential candidates with great affinity to spike protein of SARS-CoV-2 were selected. These compounds exhibited good affinity (> 7 kcal/mol). But the fifty nanoseconds molecular dynamic simulation showed the resveratrol firmly bound to viral proteins and ACE2 receptor complex. The net binding free energy calculated by MM-PBSA also confirmed that the resveratrol-viral protein complex was stable. It is expedited from the study that resveratrol can be a potential candidate for drug development against COVID-19 [82].

Method Validation

It is a very important step in computational techniques. All the techniques have specific criteria for their validity. For example, the compounds that undergo molecular docking are selected based on the best docking score. Similarly, pharmacophore modeling has a fingerprint for its validity.

CONCLUSION

The CADD approaches are very crucial in proposing the possible drug for the treatment of COVID-19. Computational methods offered some drugs (Clocortolone, Nelfinavir, Atazanavir, Indinavir, Lopinavir) to target nsp3 of SARS-CoV-2, which is the main protein for its pathogenicity. In addition, some of the antiviral drug (Remdesivir), antifungal drug (Itraconazole), anticoagulant drug (Dabigatran), drugs used to treat HIV/AIDS (Indinavir, Nelfinavir, Saquinavir, Lopinavir, Raltegravir Ritonavir, Darunavir, Dolutegravir, Efavirenz, and Atazanavir), CMV (Ganciclovir), HBV (Entecavir), HSV (Penciclovir) and HCV (Simeprevir, Ribavirin, Lomibuvir, Asunaprevir, and Grazoprevir) can provide some relief against COVID-19. There is a lack of direct clinical evidence for the efficacy of the majority of the drugs against COVID-19. However, clinical studies can conclude the most useful drug to inhibit SARS-CoV-2 progression in the future.

CONSENT FOR PUBLICATION

Not Applicable.

CONFLICT OF INTEREST

The author confirms that this chapter contents have no conflict of interest.

ACKNOWLEDGEMENT

Declared none.

REFERENCES

[1]	Myers S, Baker A. Drug discovery--an operating model for a new era. Nat Biotechnol 2001; 19(8): 727-30.[http://dx.doi.org/10.1038/90765] [PMID: 11479559]
[2]	Moses H, III, Dorsey ER, Matheson DH, Thier SO. Financial anatomy of biomedical research. JAMA 2005; 294(11): 1333-42.[http://dx.doi.org/10.1001/jama.294.11.1333] [PMID: 16174691]
[3]	Shekhar C. In silico pharmacology: computer-aided methods could transform drug development. Chem Biol 2008; 15(5): 413-4.[http://dx.doi.org/10.1016/j.chembiol.2008.05.001] [PMID: 18482690]
[4]	Song CM, Lim SJ, Tong JC. Recent advances in computer-aided drug design. Brief Bioinform 2009; 10(5): 579-91.[http://dx.doi.org/10.1093/bib/bbp023] [PMID: 19433475]
[5]	Jorgensen WL. The many roles of computation in drug discovery. Science 2004; 303(5665): 1813-8.[http://dx.doi.org/10.1126/science.1096361] [PMID: 15031495]
[6]	Zhang S. Computer-aided drug discovery and development Drug Des Discov 201123-38.
[7]	Tan JJ, Cong XJ, Hu LM, Wang CX, Jia L, Liang XJ. Therapeutic strategies underpinning the development of novel techniques for the treatment of HIV infection. Drug Discov Today 2010; 15(5-6): 186-97.[http://dx.doi.org/10.1016/j.drudis.2010.01.004] [PMID: 20096804]
[8]	Acharya C, Coop A, E, Polli J, D, MacKerell A. Recent advances in ligand-based drug design: relevance and utility of the conformationally sampled pharmacophore approach. Curr Comput Aided Drug Des 2011; 7(1): 10-22.[http://dx.doi.org/10.2174/157340911793743547]
[9]	Lauro G, Terracciano S, Cantone V, et al. A Combinatorial Virtual Screening Approach Driving the Synthesis of 2,4-Thiazolidinedione-Based Molecules as New Dual mPGES-1/5-LO Inhibitors. ChemMedChem 2020; 15(6): 481-9.[http://dx.doi.org/10.1002/cmdc.201900694] [PMID: 32022480]
[10]	Kouman KC, Keita M, Kre N’Guessan R, et al. Structure-based design and in silico screening of virtual combinatorial library of benzamides inhibiting 2-trans enoyl-acyl carrier protein reductase of Mycobacterium tuberculosis with favorable predicted pharmacokinetic profiles. Int J Mol Sci 2019; 20(19): 4730.[http://dx.doi.org/10.3390/ijms20194730] [PMID: 31554227]
[11]	Ahmed A, Choo H, Cho YS, Park W-K, Pae AN. Identification of novel serotonin 2C receptor ligands by sequential virtual screening. Bioorg Med Chem 2009; 17(13): 4559-68.[http://dx.doi.org/10.1016/j.bmc.2009.05.003] [PMID: 19464901]
[12]	Hilgenfeld R. From SARS to MERS: crystallographic studies on coronaviral proteases enable antiviral drug design. FEBS J 2014; 281(18): 4085-96.[http://dx.doi.org/10.1111/febs.12936] [PMID: 25039866]
[13]	Kandeel M, Altaher A, Alnazawi M. Molecular dynamics and inhibition of mers cov papain-like protease by small molecule imidazole and aminopurine derivatives. Lett Drug Des Discov 2019; 16(5): 584-91.[http://dx.doi.org/10.2174/1570180815666180918161922]
[14]	Zumla A, Chan JF, Azhar EI, Hui DS, Yuen K-Y. Coronaviruses - drug discovery and therapeutic options. Nat Rev Drug Discov 2016; 15(5): 327-47.[http://dx.doi.org/10.1038/nrd.2015.37] [PMID: 26868298]
[15]	Kandeel M, Al-Nazawi M. Virtual screening and repurposing of FDA approved drugs against COVID-19 main protease. Life Sci 2020; 251: 117627.[http://dx.doi.org/10.1016/j.lfs.2020.117627] [PMID: 32251634]
[16]	Gentile D, Patamia V, Scala A, Sciortino MT, Piperno A, Rescifina A. Putative inhibitors of SARS-CoV-2 main protease from a library of marine natural products: A virtual screening and molecular modeling study. Mar Drugs 2020; 18(4): 225.[http://dx.doi.org/10.3390/md18040225] [PMID: 32340389]
[17]	Wang J, Ge Y, Xie X-Q. Development and testing of druglike screening libraries. J Chem Inf Model 2019; 59(1): 53-65.[http://dx.doi.org/10.1021/acs.jcim.8b00537] [PMID: 30563329]
[18]	Friesner RA, Banks JL, Murphy RB, et al. Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. J Med Chem 2004; 47(7): 1739-49.[http://dx.doi.org/10.1021/jm0306430] [PMID: 15027865]
[19]	Wang J, Morin P, Wang W, Kollman PA. Use of MM-PBSA in reproducing the binding free energies to HIV-1 RT of TIBO derivatives and predicting the binding mode to HIV-1 RT of efavirenz by docking and MM-PBSA. J Am Chem Soc 2001; 123(22): 5221-30.[http://dx.doi.org/10.1021/ja003834q] [PMID: 11457384]
[20]	Wang J, Kang X, Kuntz ID, Kollman PA. Hierarchical database screenings for HIV-1 reverse transcriptase using a pharmacophore model, rigid docking, solvation docking, and MM-PB/SA. J Med Chem 2005; 48(7): 2432-44.[http://dx.doi.org/10.1021/jm049606e] [PMID: 15801834]
[21]	Zhou Y, Hou Y, Shen J, Huang Y, Martin W, Cheng F. Network-based drug repurposing for novel coronavirus 2019-nCoV/SARS-CoV-2. Cell Discov 2020; 6(1): 14.[http://dx.doi.org/10.1038/s41421-020-0153-3] [PMID: 33723226]
[22]	Pant S, Singh M, Ravichandiran V, Murty U, Srivastava HK. Peptide-like and small-molecule inhibitors against Covid-19. J Biomol Struct Dyn 20201-10.[PMID: 32306822]
[23]	ul Qamar MT, Alqahtani SM, Alamri MA, Chen L-L. Structural basis of SARS-CoV-2 3CLpro and anti-COVID-19 drug discovery from medicinal plants. J Pharm Anal 2020.
[24]	Lengauer T, Rarey M. Computational methods for biomolecular docking. Curr Opin Struct Biol 1996; 6(3): 402-6.[http://dx.doi.org/10.1016/S0959-440X(96)80061-3] [PMID: 8804827]
[25]	Gray JJ, Moughon S, Wang C, et al. Protein-protein docking with simultaneous optimization of rigid-body displacement and side-chain conformations. J Mol Biol 2003; 331(1): 281-99.[http://dx.doi.org/10.1016/S0022-2836(03)00670-3] [PMID: 12875852]
[26]	Van Dijk M, Bonvin AM. A protein–DNA docking benchmark. Nucleic Acids Res 2008; 26(14): e88.[http://dx.doi.org/10.1093/nar/gkn386]
[27]	Hui DS. IAE, Madani T, Ntoumi F, Koch R, Dar O. The continuing 2019-nCoV epidemicthreatof novel coronaviruses to global health: the latest 2019 novel coronavirus outbreak in Wuhan, China. Int J Infect Dis 2020; 91: 264-6.[http://dx.doi.org/10.1016/j.ijid.2020.01.009] [PMID: 31953166]
[28]	Xu X, Chen P, Wang J, et al. Evolution of the novel coronavirus from the ongoing Wuhan outbreak and modeling of its spike protein for risk of human transmission. Sci China Life Sci 2020; 63(3): 457-60.[http://dx.doi.org/10.1007/s11427-020-1637-5] [PMID: 32009228]
[29]	Liu X, Zhang B, Jin Z, Yang H, Rao Z. The crytal structure of 2019-nCoV main protease in complex with an inhibitor N3 2020.
[30]	Anand K, Ziebuhr J, Wadhwani P, Mesters JR, Hilgenfeld R. Coronavirus main proteinase (3CLpro) structure: basis for design of anti-SARS drugs. Science 2003; 300(5626): 1763-7.[http://dx.doi.org/10.1126/science.1085658] [PMID: 12746549]
[31]	Bacha U, Barrila J, Velazquez-Campoy A, Leavitt SA, Freire E. Identification of novel inhibitors of the SARS coronavirus main protease 3CLpro. Biochemistry 2004; 43(17): 4906-12.[http://dx.doi.org/10.1021/bi0361766] [PMID: 15109248]
[32]	Fiser A, Feig M, Brooks CL, III, Sali A. Evolution and physics in comparative protein structure modeling. Acc Chem Res 2002; 35(6): 413-21.[http://dx.doi.org/10.1021/ar010061h] [PMID: 12069626]
[33]	Bonneau R, Baker D. Ab initio protein structure prediction: progress and prospects. Annu Rev Biophys Biomol Struct 2001; 30(1): 173-89.[http://dx.doi.org/10.1146/annurev.biophys.30.1.173] [PMID: 11340057]
[34]	Domingues FS, Lackner P, Andreeva A, Sippl MJ. Structure-based evaluation of sequence comparison and fold recognition alignment accuracy. J Mol Biol 2000; 297(4): 1003-13.[http://dx.doi.org/10.1006/jmbi.2000.3615] [PMID: 10736233]
[35]	Marti-Renom MA, Madhusudhan MS, Fiser A, Rost B, Sali A. Reliability of assessment of protein structure prediction methods. Structure 2002; 10(3): 435-40.[http://dx.doi.org/10.1016/S0969-2126(02)00731-1] [PMID: 12005441]
[36]	Sánchez R, Šali A. Advances in comparative protein-structure modelling. Curr Opin Struct Biol 1997; 7(2): 206-14.[http://dx.doi.org/10.1016/S0959-440X(97)80027-9] [PMID: 9094331]
[37]	Fiser A, Sali A. MODELLER: generation and refinement of homology models. Methods Enzymol 2001567-89.[PMID: 14696385]
[38]	Chothia C, Lesk AM. The relation between the divergence of sequence and structure in proteins. EMBO J 1986; 5(4): 823-6.[http://dx.doi.org/10.1002/j.1460-2075.1986.tb04288.x] [PMID: 3709526]
[39]	Lesk AM, Chothia C. How different amino acid sequences determine similar protein structures: the structure and evolutionary dynamics of the globins. J Mol Biol 1980; 136(3): 225-70.[http://dx.doi.org/10.1016/0022-2836(80)90373-3] [PMID: 7373651]
[40]	Bartlam M, Yang H, Rao Z. Structural insights into SARS coronavirus proteins. Curr Opin Struct Biol 2005; 15(6): 664-72.[http://dx.doi.org/10.1016/j.sbi.2005.10.004] [PMID: 16263266]
[41]	Ge X-Y, Li J-L, Yang X-L, et al. Isolation and characterization of a bat SARS-like coronavirus that uses the ACE2 receptor. Nature 2013; 503(7477): 535-8.[http://dx.doi.org/10.1038/nature12711] [PMID: 24172901]
[42]	Li F, Li W, Farzan M, Harrison SC. Structure of SARS coronavirus spike receptor-binding domain complexed with receptor. Science 2005; 309(5742): 1864-8.[http://dx.doi.org/10.1126/science.1116480] [PMID: 16166518]
[43]	Walls AC, Park Y-J, Tortorici MA, Wall A, McGuire AT, Veesler D. Structure, function, and antigenicity of the SARS-CoV-2 spike glycoprotein. Cell 2020; 181(2): 281-292.e6.[http://dx.doi.org/10.1016/j.cell.2020.11.032]
[44]	Benvenuto D, Giovanetti M, Ciccozzi A, Spoto S, Angeletti S, Ciccozzi M. The 2019-new coronavirus epidemic: Evidence for virus evolution. J Med Virol 2020; 92(4): 455-9.[http://dx.doi.org/10.1002/jmv.25688] [PMID: 31994738]
[45]	Lu R, Zhao X, Li J, et al. Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding. Lancet 2020; 395(10224): 565-74.[http://dx.doi.org/10.1016/S0140-6736(20)30251-8] [PMID: 32007145]
[46]	Saikatendu KS, Joseph JS, Subramanian V, et al. Structural basis of severe acute respiratory syndrome coronavirus ADP-ribose-1′'-phosphate dephosphorylation by a conserved domain of nsP3. Structure 2005; 13(11): 1665-75.[http://dx.doi.org/10.1016/j.str.2005.07.022] [PMID: 16271890]
[47]	Angeletti S, Benvenuto D, Bianchi M, Giovanetti M, Pascarella S, Ciccozzi M. COVID-2019: The role of the nsp2 and nsp3 in its pathogenesis. J Med Virol 2020; 92(6): 584-8.[http://dx.doi.org/10.1002/jmv.25719] [PMID: 32083328]
[48]	Hammond RG, Tan X, Johnson MA. SARS-unique fold in the Rousettus bat coronavirus HKU9. Protein Sci 2017; 26(9): 1726-37.[http://dx.doi.org/10.1002/pro.3208] [PMID: 28580734]
[49]	Chang Y-C, Tung Y-A, Lee K-H, Chen T-F, Hsiao Y-C, Chang H-C. Potential therapeutic agents for COVID-19 based on the analysis of protease and RNA polymerase docking 2020.
[50]	Wermuth C-G, Ganellin C, Lindberg P, Mitscher L. Glossary of terms used in medicinal chemistry (IUPAC Recommendations 1998). Pure Appl Chem 1998; 70(5): 1129-43.[http://dx.doi.org/10.1351/pac199870051129]
[51]	Gund P. Evolution of the Pharmacophore Concept in 2000.
[52]	McGregor MJ, Muskal SM. Pharmacophore fingerprinting. 1. Application to QSAR and focused library design. J Chem Inf Comput Sci 1999; 39(3): 569-74.[http://dx.doi.org/10.1021/ci980159j] [PMID: 10361729]
[53]	McGregor MJ, Muskal SM. Pharmacophore fingerprinting. 2. Application to primary library design. J Chem Inf Comput Sci 2000; 40(1): 117-25.[http://dx.doi.org/10.1021/ci990313h] [PMID: 10661558]
[54]	Mason JS, Morize I, Menard PR, Cheney DL, Hulme C, Labaudiniere RF. New 4-point pharmacophore method for molecular similarity and diversity applications: overview of the method and applications, including a novel approach to the design of combinatorial libraries containing privileged substructures. J Med Chem 1999; 42(17): 3251-64.[http://dx.doi.org/10.1021/jm9806998] [PMID: 10464012]
[55]	Geppert TD, Lipsky PE. Antigen presentation at the inflammatory site. Crit Rev Immunol 1989; 9(4): 313-62.[PMID: 2679663]
[56]	Sheridan RP, Rusinko A, III, Nilakantan R, Venkataraghavan R. Searching for pharmacophores in large coordinate data bases and its use in drug design. Proc Natl Acad Sci USA 1989; 86(20): 8165-9.[http://dx.doi.org/10.1073/pnas.86.20.8165] [PMID: 2813386]
[57]	Jones G, Willett P, Glen RC. A genetic algorithm for flexible molecular overlay and pharmacophore elucidation. J Comput Aided Mol Des 1995; 9(6): 532-49.[http://dx.doi.org/10.1007/BF00124324] [PMID: 8789195]
[58]	Wolber G, Seidel T, Bendix F, Langer T. Molecule-pharmacophore superpositioning and pattern matching in computational drug design. Drug Discov Today 2008; 13(1-2): 23-9.[http://dx.doi.org/10.1016/j.drudis.2007.09.007] [PMID: 18190860]
[59]	Hastings WK. Monte carlo sampling methods using markov chains and their applications. Biometrika 1970; 57(1): 97-109.[http://dx.doi.org/10.1093/biomet/57.1.97]
[60]	Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E. Equation of state calculations by fast computing machines. J Chem Phys 1953; 21(6): 1087-92.[http://dx.doi.org/10.1063/1.1699114]
[61]	Ponder JW, Case DA. Force fields for protein simulations Advances in protein chemistry 200327-85.
[62]	Frenkel D, Smit B. Understanding molecular simulation: From algorithms to applications. Elsevier (formerly published by Academic Press), 2002; p. 1-638
[63]	Paquet E, Viktor HL. Molecular dynamics, monte carlo simulations, and langevin dynamics: a computational review. Biomed Res Int 2015183918.[http://dx.doi.org/10.1155/2015/183918]
[64]	Andreani J, Guerois R. Evolution of protein interactions: from interactomes to interfaces. Arch Biochem Biophys 2014; 554: 65-75.[http://dx.doi.org/10.1016/j.abb.2014.05.010] [PMID: 24853495]
[65]	Petta I, Lievens S, Libert C, Tavernier J, De Bosscher K. Modulation of protein–protein interactions for the development of novel therapeutics. Mol Ther 2016; 24(4): 707-18.[http://dx.doi.org/10.1038/mt.2015.214] [PMID: 26675501]
[66]	Rodrigues JP, Bonvin AM. Integrative computational modeling of protein interactions. FEBS J 2014; 281(8): 1988-2003.[http://dx.doi.org/10.1111/febs.12771] [PMID: 24588898]
[67]	Kaczor AA, Selent J, Poso A. Structure-based molecular modeling approaches to GPCR oligomerization Methods in cell biology 201391-104.
[68]	Kuntz ID, Blaney JM, Oatley SJ, Langridge R, Ferrin TE. A geometric approach to macromolecule-ligand interactions. J Mol Biol 1982; 161(2): 269-88.[http://dx.doi.org/10.1016/0022-2836(82)90153-X] [PMID: 7154081]
[69]	Park H, Lee H, Seok C. High-resolution protein-protein docking by global optimization: recent advances and future challenges. Curr Opin Struct Biol 2015; 35: 24-31.[http://dx.doi.org/10.1016/j.sbi.2015.08.001] [PMID: 26295792]
[70]	Kaczor AA, Selent J, Sanz F, Pastor M. Modeling complexes of transmembrane proteins: systematic analysis of protein-protein docking tools. Mol Inform 2013; 32(8): 717-33.[http://dx.doi.org/10.1002/minf.201200150] [PMID: 27480064]
[71]	Zacharias M. Accounting for conformational changes during protein-protein docking. Curr Opin Struct Biol 2010; 20(2): 180-6.[http://dx.doi.org/10.1016/j.sbi.2010.02.001] [PMID: 20194014]
[72]	Ban T, Ohue M, Akiyama Y. Multiple grid arrangement improves ligand docking with unknown binding sites: Application to the inverse docking problem. Computational biology and chemistry 2018; 73: 139-46.[http://dx.doi.org/10.1016/j.compbiolchem.2018.02.008]
[73]	Kapetanovic IM. Computer-aided drug discovery and development (CADDD): in silico-chemico-biological approach. Chem Biol Interact 2008; 171(2): 165-76.[http://dx.doi.org/10.1016/j.cbi.2006.12.006] [PMID: 17229415]
[74]	Wu L, Shen Y, Li M, Wu F-X. Network output controllability-based method for drug target identification. IEEE T NANOBIOSCI 2015; 14(2): 184-91.[http://dx.doi.org/10.1109/TNB.2015.2391175] [PMID: 25643411]
[75]	Wishart DS, Knox C, Guo AC, et al. DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res 2008; 36(Database issue) (Suppl. 1): D901-6.[http://dx.doi.org/10.1093/nar/gkm958] [PMID: 18048412]
[76]	Cheng AC, Coleman RG, Smyth KT, et al. Structure-based maximal affinity model predicts small-molecule druggability. Nat Biotechnol 2007; 25(1): 71-5.[http://dx.doi.org/10.1038/nbt1273] [PMID: 17211405]
[77]	Keiser MJ, Roth BL, Armbruster BN, Ernsberger P, Irwin JJ, Shoichet BK. Relating protein pharmacology by ligand chemistry. Nat Biotechnol 2007; 25(2): 197-206.[http://dx.doi.org/10.1038/nbt1284] [PMID: 17287757]
[78]	Beck BR, Shin B, Choi Y, Park S, Kang K. Predicting commercially available anti-viral drugs that may act on the novel coronavirus (SARS-CoV-2) through a drug-target interaction deep learning model. Comput Struct Biotechnol J 2020; 18: 784-90.
[79]	Alamri MA. Pharmacoinformatics and molecular dynamic simulation studies reveal potential inhibitors of SARS-CoV-2 main protease 3CLpro. J Biomol Struct Dyn 2020; 39(13): 4936-4948
[80]	Kumar D, Kumari K, Jayaraj A, Kumar V, Kumar RV, Dass SK. Understanding the binding affinity of noscapines with protease of SARS-CoV-2 for COVID-19 using MD simulations at different temperatures. J Biomol Struct Dyn 20201-14.[http://dx.doi.org/10.1080/07391102.2020.1752310] [PMID: 32362235]
[81]	Ibrahim IM, Abdelmalek DH, Elshahat ME, Elfiky AA. COVID-19 spike-host cell receptor GRP78 binding site prediction. J Infect 2020; 80(5): 554-62.[http://dx.doi.org/10.1016/j.jinf.2020.02.026] [PMID: 32169481]
[82]	Wahedi HM, Ahmad S, Abbasi SW. Stilbene-based natural compounds as promising drug candidates against COVID-19. J Biomol Struct Dyn 2020 (just-accepted): 1-16