Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5345_Библиотеки_им_академика_М_И_Перельмана.pdf
Integrated in Silico Methods for the Design and Optimization of Novel Drug Candidates
Zupan, J., Novič, M., & Ruisánchez, I. (1997). Kohonen and counterpropagation artificial neural networks
in analytical chemistry. Chemometrics and Intelligent Laboratory Systems, 38(1), 1–23. doi:10.1016/
S0169-7439(97)00030-0
KEY TERMS AND DEFINITIONS
Antibacterial Agents: A common term for chemicals including synthetic or semi-synthetic drugs
or other similar chemical entities that either kill or inhibit the bacterial growth.
Artificial Neural Networks: Computational non-linear modeling tools that mimic the structure and
functions of the biological neural networks in the brain. During the learning process (training of the
network), a complex relationship between the input and output data is established. They can be used
either for numerical predictions of properties or pattern recognition purposes.
DNA Gyrase: An omnipresent molecular nanomachine from the type II DNA topoisomerase super-
family that is responsible for the unwinding of the DNA molecule during the DNA replication phase.
The bacterial DNA gyrase is a well-established target of many antibacterials including nalidixic acid
and their derivatives 6-fluoroquinolones.
Drug-Likeness: A qualitative measure based on a set of complex in silico calculable physico-chemical
properties (e.g., molecular weight, logP, number of rotatable bonds, number of hydrogen bond donors
and acceptors, and polar surface area) that determine whether an investigated compound is similar to the
known drugs. It was found as a useful measure in the modern drug discovery and it is frequently used
to filter out the so-called “drug-like” compounds from massive chemical libraries.
Protein Homology Modeling: Construction of a three-dimensional model of the “target” protein at
atomic resolution from its amino acid sequence and an experimental three-dimensional structure of a
related homologous protein (“template”).
Quantitative Structure-Activity Relationship: An approach designed to establish relationships
between the chemical structure and biological activity (or other target property) of investigated compounds in a quantitative manner.
Virtual Combinatorial Library Design: Generation of a list of structurally similar molecules in
a virtual (in silico) environment employing the principles of combinatorial chemistry where a set of
reagents (substituents or building-blocks) are specifically attached at pre-defined scaffold positions on
the main structure.
Virtual Screening: A computational methodology used in the modern drug discovery to search and
identify those molecular entities (small molecules) from a chemical library which are most likely to bind
to a drug target, usually protein receptor or enzyme.
316
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
Integrated in Silico Methods for the Design and Optimization of Novel Drug Candidates
ENDNOTES
1
http://www.knime.org
2
http://accelrys.com/products/pipeline-pilot
3
http://sourceforge.net/apps/mediawiki/cdk/index/.php?title=Main_Page
4
http://www.rdkit.org
5
http://ggasoftware.com/opensource
6
http://www.novamechanics.com/index.php
7
https://github.com/knime-mpicbg/HCS-Tools/wiki
8
http://www.seqan.de
9
https://www.molecular-networks.com/pipelinepilot
10
http://www.biosolveit.de/PipelinePilot
11
http://www.cresset-group.com/cresset-pipeline-pilot-component-v2-0-0-release-notes
12
http://cdktaverna.wordpress.com
13
http://www.inhibox.com
14
https://www.sciencecloud.com
15
http://accelrys.com/products/databases/sourcing/available-chemicals-directory.html
16
http://accelrys.com/products/discovery-studio/pharmacophore-ligand-based-design.html
17
http://www.inteligand.com/ligandscout
18
http://chemdb.niaid.nih.gov
19
http://www.keyorganics.ltd.uk
20
https://www.ccdc.cam.ac.uk/Solutions/GoldSuite/pages/GoldSuite.aspx
21
http://www.asinex.com
22
http://www.eyesopen.com/rocs.
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
317
318
Chapter 9
Computational Approaches
for the Discovery of Novel
Hepatitis C Virus NS3/4A
and NS5B Inhibitors
Khac-Minh Thai
University of Medicine and Pharmacy at HCMC,
Vietnam
Quoc-Hiep Dong
University of Medicine and Pharmacy at HCMC,
Vietnam
Thi-Thanh-Lan Nguyen
University of Medicine and Pharmacy at HCMC,
Vietnam
ABSTRACT
Nonstructural 5B (NS5B) polymerase and Nonstructural 3/4A (NS3/4A) protease have proven to be
promising targets for the development of anti-HCV (Hepatitis C Virus) agents. The NS5B polymerase is
of paramount importance in HCV viral replication; therefore, employing NS5B inhibitors was considered an effective way for the treatment of HCV. Identifying inhibitors against NS3/4A serine protease
represents another attractive approach applied in anti-HCV drug discovery, which is evidenced by its
crucial role of in the biogenesis of the viral replication activity. In this chapter, many different computational approaches including Quantitative Structure-Activity Relationship (QSAR) and virtual screening
in anti-HCV drug discovery were considered and discussed in detail. Virtual Screening (VS) techniques,
including ligand-based and structure-based, and QSAR have been utilized for the discovery of NS5B
inhibitors. Moreover, using various in silico protocols and workflows, a number of studies have been
conducted with an aim of identifying potential NS3/4A blockage agents.
University of Medicine and Pharmacy at HCMC,
University of Medicine and Pharmacy at HCMC,
University of Medicine and Pharmacy at HCMC,
Duy-Phong Le
Vietnam
Minh-Tri Le
Vietnam
Thanh-Dao Tran
Vietnam
DOI: 10.4018/978-1-4666-8136-1.ch009
Copyright © 2015, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
INTRODUCTION
Hepatitis C virus infection causes a global healthcare burden, which is likely to increase in the coming
years. There are approximately 3 to 4 million new cases of HCV infection each year, and it is estimated
that a minimum of 3% of worldwide population are chronically infected (Wasley et al., 2000). In most
infected patients, this remarkable RNA virus causes the development of malignant chronic diseases, including cirrhosis and hepatocellular carcinoma, which often leads to liver failure and death (Alter, 1993;
Hoofnagle, 1997). At present, neither an HCV vaccine nor an effective therapy against all genotypes
of HCV is available. The current therapy, including pegylated interferon α, either alone or in combination with ribavirin, a broad spectrum antiviral agent, has not only limited efficacy, but also significant
adverse effects (Feld et al., 2005; Fried, 2002). Hence, it is essential to develop novel agents with a high
therapeutic efficacy, reduced side effects, and convenient administration to meet requirements for an
anti-HCV agent.
Perceiving that urgent need, scientists around the world have conducted various research to identify as many novel HCV antiviral agents as possible. Currently, different targets for HCV therapeutic
intervention encompass both structural and non-structural proteins. The structural protein is processed
by host and viral proteases into four structural (core, E1, E2, and p7) and six nonstructural proteins
(NS2, -3, -4A, -4B, -5A, and -5B) (Shimakami et al., 2009; Tanji et al., 1994). Non-structural protein
3/4A (serine protease - helicase) and non-structural protein 5B (RNA-dependent RNA polymerase RdRp) have attracted the attention of medicinal chemists as targets for drug development because they
play a vital role in HCV replication and the host lacks functional counterparts of them (Shimakami et
al., 2009; Wang et al., 2000). This chapter will generally point out certain conducted computational
approaches and virtual screening, which could be employed in combination with QSAR to build a
robust model, with the goal of identifying new effective antiviral drugs utilizing NS3/4A and NS5B
as principle targets.
HCV GENOME AND STRUCTURE
HCV Genome
HCV is an enveloped, positive-sense and single-stranded RNA virus approximately 9600 nucleotides
in length. The significant genetic diversity was exhibited in HCV genome due to its highly error prone
RNA polymerase, which makes difficulties for vaccine development and the discovery of anti–HCV
agents (Francesco et al., 2005). 6 major HCV genotypes have been identified with the difference of
over 30% in nucleotide sequences among each of them (Simmonds et al., 2005). RNA of HCV includes
one continuous open reading frame bounded by two nontranslated regions (NTRs) at 5’ and 3’ ends.
5’ NTR constitutes the internal ribosome entry site (IRES), which plays the role as a starting point of
cap-independent translation of HCV genome to produce a single polyprotein (Honda et al., 1996). The
polyprotein is in turn split by both host cell and virus proteases into 10 different viral proteins with various functions and characteristics (Alvisi et al., 2011).
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
319
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
Structural Proteins
HCV core contains a nucleocapsid protein which fulfills numerous important functions involving RNA
binding, immune modulation, signaling, and autophagy. The core protein is surrounded by HCV E1/
E2 which are glycosylated envelope glycoproteins. Additionally, glycosylated proteins are able to neutralize antibody, which is attributed to ineffective immune and HCV persistence (Kaplan et al., 2007).
Furthermore, there is another ion channel protein, namely p7, which is of great importance in viral assembly and release.
Non-Structural Proteins
The non-structural protein (NS) NS2 acts as the viral auto-protease, which induces the cleavage
between NS2 and NS3. NS3 plays two distinct roles, as serine protease at N-terminal region and
as RNA helicase at C-terminal one. NS3 protease displays a crucial function in HCV replication
system by splitting downstream of NS3 at 4 site (between NS3/4a, NS4A/4B, NS4B/5A, NS5A/5B).
Moreover, the TLR3 adaptor protein TRIF and mitochondrial antiviral signaling protein MAVS
are also cleaved by NS3 protease, these cleavages also prevent the induction of cellular type 1 IFN
pathway (Garcia-Sastre et al., 2006). In order to perform its tasks, NS3 protease requires cofactor
NS4A. The function of NS4B has not been fully understood, probably associated with the formation of membranous web. The RNA viral and various host factors are bound together in a position
close to HCV core by NS5A, which is a dimeric zinc-binding metalloprotein. On the other hand,
NS5B is the RNA-dependent RNA polymerase (RdRp), which mediates viral genome replication
(Chang et al., 2013).
COMPUTATIONAL METHOD IN DRUG DESIGN
The following part aims at introducing several concepts of computational method. QSAR is discussed
in detail somewhere of this book, so, in this part, we only give a brief introduction of QSAR and mainly
focus on virtual screening including pharmacophore and docking.
QSAR Modeling
Quantitative structure-activity relationship (QSAR) is a computational method which uses diverse statistical and mathematical methods to identify the correlation between bioactivity and molecular descriptors. QSAR has various applications including bioactivity, selectivity, ADMET prediction (absorption,
distribution, metabolism, exertion, and toxicity). Consequently, QSAR helps medical chemists identify
active or selective compounds and increase the knowledge of SAR. By carrying out QSAR study, the
number of synthesized compounds would be decreased, and least possible in vitro or in vivo experiments
need to be performed. Fundamentally, QSAR model building process could be divided into three steps:
320
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
1. Extraction of Molecular descriptors from structures,
2. Selection of relevant molecular descriptors, and
3. Generation of QSAR model. (Duke et al., 2006)
3D Pharmacophore Modeling
Paul Ehrlich was the first man to provide the definition of pharmacophore as common molecular features
of ligands (Drie et al., 2007). The present concept, however, was introduced by Wemuth et al. (Wemuth
et al., 1998) and was employed as IUPAC definition of pharmacophore:
A pharmacophore is the ensemble of steric and electronic features that is necessary to ensure the optimal
supramolecular interactions with a specific biological target and to trigger or block its biological response.
3D pharmacophore modeling is a reputed in silico technique with various advantages in drug discovery. Firstly, pharmacophore-based models provide chemical features that are supposed critical for
interaction between ligand and receptor. Secondly, pharmacophore-based screening is capable of filtering out millions of compounds in reasonable time to identify a shortlist of hit compounds with various
biosteric scaffolds. A significant number of books, book chapters, and reviews on pharmacophore-based
virtual screening could be found in the literature, in which the most recent one is by Leach et al. (Leach
et al., 2010).
Pharmacophore-based screening is a virtual screening approach that can be applied to a large number
of compunds (Walter et al., 1998). It is basically used to retrieve novel lead structures from commercial
or synthesizable structure databases. In addition to screening application, new biological activities or
side effects are capable of being predicted by pharmacophore-based models. With the assistance of
pharmacophore-based screening, high-throughput screening provides remarkably higher enrichment
of active compounds, for instance, a study was conducted by Doman et al (Doman et al., 2002) which
exhibited a promising result. Furthermore, integrating pharmacophore model into QSAR is another
auspicious approach to generate robust models and to give elevated enrichment of true ligands.
Fundamentally, a standard pharmacophore-generating procedure includes four steps:
1. Selection of a set of ligands (only in ligand-based pharmacophore)
2. Conformational analysis,
3. Pharmacophore model generation, and
4. Pharmacophore model evaluation.
In the first step, a set of high potent ligand is selected. Then, 3D structure of ligands employed in
building pharmacophore model should exhibit bioactive conformation which is able to be extracted from
co-crystallographic structures. There is, however, a limited number of co-crystallographic structures available in protein data bank (PDB) website. It is abundantly evidenced that a set of diverse local minimized
conformers is another effective representation of the bioactive conformers. Following that, based on the
bioactive conformers of the ligands, a pharmacophore model is generated either by manual or elucidation algorithm. There are 6 kinds of model feature in pharmacophore involving hydrophobic centroids,
aromatic rings, hydrogen bond acceptors or donors, cations, and anions. Manifold commercial software
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
321
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
packages are available, such as DS CATALYST, LIGANDSCOUD, PHASE, and MOE. Finally, a new
pharmacophore model is evaluated by sets of active or inactive compounds with an aim of qualifying
this enrichment of active compounds. Hopefully, active compounds with diverse scaffolds will match
pharmacophore model features as much as possible, and inactives will not be retrieved by the model.
The enrichment of pharmacophore models is assessed by two basic values, namely the sensitivity (Se)
and the specificity (Sp). (Triballeau et al., 2005).
selected actives
Se
= =
all actives
discarded intives
Sp
= =
all inactives
TP
+
TP FN
TN
+
TN FP
• TP: True positive (these compounds are truly identified as positive).
• TN: True negative (these compounds are truly identified as negative).
• FP: False positive (these compounds are falsely identified as positive).
• FN: False negative (these compounds are falsely identified as negative).
Depending on the objects, pharmacophore modeling is divided into two different approaches, namely
structure-based and ligand-based pharmacophore.
Structure-Based Pharmacophore Modeling
The approach employs the information from protein-ligand interaction achieved from 3D co-crystallographic structure to generate models. Although it remains many limitations, especially protein flexibility, structure-based pharmacophore have gained more and more interest due to the growing number of
protein-ligand complex crystal structures. In structure-based pharmacophore, model features usually are
derived from protein-ligand interactions by using heuristics (LIGANDSCOUT) (Wolber at al., 2005),
force field energy calculation in GBPM (Grid-based pharmacophore model) (Ortuso et al., 2006), or
pharmacophore finger-print in FLIP (Karnachi et al., 2006).
Ligand-Based Pharmacophore
In case the co-crystallographic structure of ligand-protein complex is unavailable, known potent
ligands could be used to generate pharmacophore model. In this approach, either common chemical
features are identified by superimposing conformers of training set of active compounds used to
build a model or a single high potent ligand is employed as a template for generating pharmacophore
manually. Numerous commercial packages are available such as DS CATALYST, GALAHAD,
GASP, LIGANDSCOUD, MOE, and PHASE to construct pharmacophore models. Additionally,
there are two novel ligand-based pharmacophore approaches including pharmacophore fingerprint,
by Young et al. (Young et al., 2006) and feature tree, by Rarey et al. (Rarey et al., 1998), which
have given promising results.
322
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
Docking
Docking, a virtual screening method, includes the prediction of ligand conformations and orientation
within a protein binding site and the estimation of ligand-protein interaction energy. Fundamentally,
docking process involve two distinct steps, namely docking and scoring.
In the first step, a conformational search algorithm is employed to simulate all suitable conformers
of a ligand which fit into the active site of the protein. With an aim of performing conformational search
task, some software packages, for instance, DOCK (Kuntz et al., 1994), CAESAR (Ehlers et al., 2007),
and OMEG (Bostrom et al., 2001) generate a set of 3D possible conformers for all ligands to be docked
(so-called flexibases) (Kearslry et al., 1994). In another approach, a method called incremental construction which splits a compound into many principle fragments is used. Based on geometry matching, a
reference fragment is docked into the binding site. Finally, the full ligand is reconstructed incrementally
by adding fragments stepwise until the whole molecule has been completed. This algorithm is implemented in many commercial software packages, such as FLEXX (Rarey et al., 1996) and SURFLEX
(Jain, 2003). In the recent development of docking, two category of stochastic methods are employed
including genetic algorithms and Monte Carlo (MC). In MC method, one parameter is varied at a time
(torsional angle and global rotation/ translation), and the energy of generated pose is calculated. If that
energy is not higher than the previous one, the conformer is kept. If energy of the conformer is higher,
the probability of that pose will be identified by a selection process based on Boltzmann weight factor
exp(-E/kT). MC method is usually accompanied by simulate annealing that traps the protein-ligand complex in a low energy minimum by decreasing temperature (in GLAMDOCK) (Tietze and Apostolakis.,
2007). The process continuously declines the freedom of the ligand until reaching global minimum.
GOLD (Jones et al., 1997) is the first docking software to use genetic algorithm (GA) in which genes
in a chromosome encode conformational properties of a ligand pose. Chromosomes are evaluated by
fitness function, and the strongest ones are allowed to produce offspring until a termination criterion is
obtained, for example RMSD within a certain threshold. Both stochastic methods probably experience a
local minimum in which global one of potential surface energy can be out of reach, however, its results
usually outperform traditional methods (Douglas et al., 2004).
The second step of docking process is to calculate binding energy by scoring function. Generally, there are
three kinds of scoring function including empirical, knowledge-based, and force field scoring function. In the
first approach, several terms, usually polar and apolar interactions, are used to describe important properties
in ligand binding in order to build a mathematic equation for predicting binding energy (Ferrara et al., 2004).
Knowledge-based method employs the sum of all interatomic interactions (Helmholtz free interaction energies
of protein-ligand atom pair) of protein-ligand complex to estimate the docking score. The most recent scoring
function is force field-based scoring functions calculating the binding energy of protein-ligand complex on the
basis of classical molecular mechanics force fields (AMBER and CHARMM). Van der Waals interaction is
described by Lennard-Jones potential, wherase the electrostatic interaction is described by Coulomb energy.
Unfortunately, all three scoring functions are not very accurate in prediction of binding energy due to the
omission of many factors, such as entropic component, solvation, and protein flexibility. Therefore, they are
not robust enough to rank hits by elevating binding energy values. Indeed, pharmacophore modeling often
outperforms docking in enrichment rate of active compound, remarkably in a very large database set (Chen
et al., 2009). However, docking remains vital in drug design as a method in combination with QSAR to constructing a robust model. Docking is able to give many important features in protein-ligand binding to choose
as descriptors for QSAR model or to re-evaluate high potential compounds with high predicted pIC
value.
50
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
323
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
NONSTRUCTURAL PROTEIN 5B (NS5B)
Structure of NS5B
Like other RNA dependent RNA polymerases, NS5B is an error-prone enzyme, having a high error rate
of its part. The viral replicase is much less accurate than the replicative prokaryotic or eucaryotic DNA
polymerase due to the lack of exonuclease and proofreading domain. This also explains the reason why
HCV has rapidly a wide variety of genotypes (Patel et al., 2008).
Twenty eight co-crystallographic structures of HCV NS5B have been determined to date (PDB
ID: 1Z4U, 1YVF, 1YVX, 1YVZ, 2DXS, 2HWH, 2HWI, 2JCI, 2JCO, 2YOJ, 2AWZ, 2AXO, 2AXI,
2BRK, 2BRL, 2I1R, 2HA1, 2D3U, 2D3Z, 2D4I, 2GC8, 3VQ5, 3PHE, 3LKH, 3CWJ, 3CDE, 3BR9,
3OD5). It revealed that the structure of NS5B is similar to the structure of polymerase of bacteriophage Ф6 because of its fully encircled-active site (Elhefnawi et al., 2012). Like other templatedependent nucleotide polymerases, NS5B enzyme is imaginatively assimilated to the right hand. It
consists of four subdomains, referred to as the palm (in violet), the thumb (in green) and the fingers
(in blue) and C-terminal domain (in yellow) illustrated in (Figure 1) (T. Li et al., 2010). The palm
domain contains the active site aspartates (Asp220, Asp318, Asp319) of the enzyme; whereas, the
fingers and the thumb are responsible for interaction with RNA chain (Musmuca et al., 2010). The
extensions of fingers and thumb lead to a more fully-enclosed active site. In the thumb domain,
there is a beta-hairpin which is proposed to move upon the formation of the existing double-stranded
RNA (Elhefnawi et al., 2012).
Figure 1. Structure of NS5B (PDB ID: 2DXS)
324
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
Computational Approaches for the Discovery of Novel Hepatitis C Virus NS3/4A and NS5B Inhibitors
Function of NS5B
NS5B plays an essential role in HCV viral replication (Shimakami et al., 2009; Tanji et al., 1994). The
key function of viral replication includes both using viral positive RNA strand as its templates and
catalyzing the polymerization of ribonucleoside triphosphates (rNTP) during RNA replication (Shimakami et al., 2009). The encircled-active site of NS5B enzyme is the aspartic acid residues involved
in the nucleotidyl transfer reaction during polymerization (Feld et al., 2005). Two divalent metal ions
2+
like Mg
and facilitated the nucleophilic attack of 3’-hydroxy group of primer terminus or priming nucleotide on
the α-phosphate of the incoming nucleotide. The ions may also be a cause of release of pyrophosphates
(PPi) product (Powdrill et al., 2010).
which is proposed to be the mechanism of initiation in vivo. To produce the (+)-strand RNA, the positive
strand RNA is used as the template to synthesize the minus strand RNA. In the initial phase, the polymerase plays a role of priming and initiating nucleotides by catalyzing the formation of a phosphodiester
bond between two bound nucleotides. The next step takes place when the newly formed dinucleotides
are used as a primer for addition of the third nucleotide, while the enzyme switches to the elongation
mode (Powdrill et al., 2010).
(Powdrill et al., 2010). Binding a structure at this active site to alter its functionality as well as intefering the function of the enzyme at each of the aforementioned individual steps by some small molecular
NS5B inhibitors can inhibit viral RNA replication (Li et al., 2010; Melagraki et al., 2011).
and Mn2+ are coordinated with the three aspartic acid carboxylates (D220, D318 and D319),
The HCV NS5B is capable of initiating RNA synthesis de novo, especially in the absence of primer,
The active site of NS5B controls possibly nucleotid binding and initiates the de novo RNA synthesis
QSAR APPROACH EMPLOYED TO IDENTIFY NOVEL NS5B INHIBITORS
2D QSAR Approach
Several research into identifying NS5B inhibitors were carried out and a number of small molecular
anti-NS5B compounds have been discovered. These inhibitors are categorized into two general types,
namely nucleoside analogs that function as chain terminators, and non-nucleoside inhibitors (NNIs) that
bind near the active site or allosteric sites on NS5B (Carroll et al., 2003; Hang et al., 2009). The NNIs
include compounds belonging to benzimidazole, indole, thiophene, phenylalanine, dihydropyranone,
pyranoindole, benzothiadiazine, proline sulfonamide, benzylidene, diketoacids, acrylic acid, tetracyclic
indole, quinoxaline, and indole N-acetamide scaffolds (Patel et al., 2008).
QSAR approach, one of ligand-based methods, was utilized in several research to discover innova-
tive NS5B inhibitors.
Melagraki et al. asserted the results of a ligand-based virtual screening on 98 compounds (benzothiadiazine analogs) which assessed as inhibitors of genotype 1 HCV polymerase (Melagraki et
al., 2007). The dataset was divided into a training set of 60 compounds, and a validation set of 38
compounds based on the Kennard and Stones algorithm. Initially, quantitative structure-activity patterns were investigated for the chosen molecules. As a result, an accurate and reliable QSAR model
involving five descriptors that is capable of predicting successfully the HCV inhibitory potency
against genotype 1 HCV polymerase was developed. These five descriptors included lipophilicity,
EBSCOhost - printed on 2/14/2023 7:16 AM via . All use subject to https://www.ebsco.com/terms-of-use
325