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_5435_Библиотеки_им_академика_М_И_Перельмана
.pdf
https://t.me/med1917

Mohammad Ovais Dar, Aamir Tariq Malla, Zahid Ahmad Paul,
Roohi Mohi ‑ud‑din, Mubashir Hussain Masoodi, Pooja A Chawla
✶
,
and Reyaz Hassan Mir
✶
12 Unlocking therapeutic potential:
computational approaches for enzyme
inhibition discovery
Abstract: Enzyme inhibition is a critical strategy in drug discovery and develop-
ment, aimed at modulating various biological processes f or therapeutic purposes.
Computational studies have emerged as i ndispensable tools in this pursuit, allowing
for efficient screening and design of potential inhibitors. This comprehensive re-
view highlights the key computational techniques utilized in the identification and
optimization of enzyme inhibitors, focusing on both lig and-based and structure-
based approaches. Ligand-based strategies include virtual screening, quantitative
structure–activity relationship modeling, and pharmacophore-based analyses, offer-
ing insights into the bioactivity and selectivity of potential inhibitors. Structure-
based methods, including molecular docking, molecular dynamics simulations, and
free energy calculations, provide valuable information about binding modes, ener-
getics, and conformational changes within enzyme-inhibitor complexes. Overall,
this work demonstrates the pivotal role of computational studies in the discovery of
enzyme inhibitors and underscores the advantages of combining ligand and struc-
ture-based drug design strategies. By integrating these approaches, researchers can
streamline the drug discovery process, optimize lead compounds, and pave the way
for the development of innovative therapies tar geting a wide range of diseases.
Keywords: drug discovery, enzyme inhibitor, drug design, molecular docking, bioactivity
✶
Corresponding author: Pooja A Chawla, University Institute of Pharmaceutical Sciences and
Research Baba Farid University of Health Sciences Sadiq Road Faridkot 151203, Punjab India,
e-mail: pvchawla@gmail.com
✶
Corresponding author: Reyaz Hassan Mir, Pharmaceutical Chemistry Division, Department of
Pharmaceutical Sciences, University of Kashmir, Srinagar 190006, Jammu and Kashmir, India,
e-mail: reyazhassan249@gmail.com
Mohammad Ovais Dar, Department of Pharmaceutical Chemistry, M. M. College of Pharmacy,
Maharishi Markandeshwar (Deemed to be University), Mullana, Ambala 133207, Haryana, India
Aamir Tariq Malla, Zahid Ahmad Paul, Mubashir Hussain Masoodi, Pharmaceutical Chemistry
Division, Department of Pharmaceutical Sciences, University of Kashmir, Hazratbal, Srinagar 190006,
Jammu and Kashmir, India
Roohi Mohi‑ud‑din, Department of General Medicine, Sher‑I‑Kashmir Institute of Medical Sciences
(SKIMS), Srinagar 190001, Jammu and Kashmir, India
https://doi.org/10.1515/9783111207117-012
https://t.me/med1917

12.1 Introduction
Leading pharmaceutical companies and research organizations have accelerated the
discovery and development process of the drug by using computer-aided drug discov-
ery (CADD) tools in preliminary investigations to reduce prices and the final stage fail-
ures [1]. Understanding the molecular interactions and the binding affinity between
the ligand and the target protein is made easier by using rational drug design as a key
component of CADD. Moreover, the advent of parallel processing, supercomputing fa-
cilities, and algorithms, improved software, and other tools has made identification of
lead easier in pharmaceutical research [2]. The understanding, analysis, and justifica-
tion of the pharmaceutical related big data in the process of drug discovery have also
been substantially facilitated by recent developments in machine learning (ML) and
artificial intelligence [3]. CADD is typically classified into two classes which include
ligand-based and structure-based approaches. These two methods have been exten-
sively employed in the process of drug discovery to find viable lead compounds. In
contrast to structure-based approach, which depends on the target receptor three-
dimensional (3D) structure and also its active sites to comprehend the molecular in-
teractions among the ligand and the receptor, the ligand-based approach relies on in-
formation about the ligands that interact with a particular receptor [4]. There are
several examples of successful CADD, and it remains a crucial component of the drug
discovery process [5].
12.2 Ligand-based drug design
The ligand-based drug design approach or indirect drug design [6] is employed when
the 3D information about the receptor is unavailable. Instead, it depends on the struc-
tural information of the ligands which bind to the receptor of interest [7]. The infor-
mation about these molecules or ligands can be utilized to create a p harmacophore
model, which identifies the minimum essential structural features required for a mol-
ecule to interact with the target receptor. Therefore, used to design enzyme inhibitors,
design of antagonists and agonists of various neurotransmitters. In other words, by
utilizing the knowledge about the binding characteristics of the target receptor, it is
possible to construct a model of the biological target, which can then be used to de-
velop new molecules that have the ability of binding to the target receptor. With the
structural knowledge of active compounds (that already binds to the target), the syn-
thetic chemist has the opportunity to design and synthesize related compounds in
order to access how their biological activity changes with structure through quantita-
tive structure–activity relationship (QSAR) studies. An advantage of the LBDD is that
no information of the drugs exact molecular-level mechanism of action is required, as
long as there is some measure of biological activity [8].
296 Mohammad Ovais Dar et al.
https://t.me/med1917

The main principle of ligand-based CADD (LB-CADD) depends on the concept of
similar properties, according to which, substances or molecules with alike structural
characteristics are likely to exhibit similar properties [9]. B y using LB-CADD tech-
nique, ligands which are confirmed to bind with the intended target are evaluated.
This process aims to organize and preserve the crucial molecular descriptors related
to the structure and physicochemistry of compounds or ligands in order to prioritize
desirable interactions with the intended target. Any unrelated interaction or informa-
tion is disregarded [10]. The basic techniques of LB-CADD consist of molecular similar-
ity-based search, pharmacophore modeling, and QSAR [11].
12.2.1 Molecular similarity-based search
The molecular similarity-based search processisthesimplestLBDDapproachtodiscover
promising molecules, and it is a crucial part of other SBDD and LBDD techniques that
involve the exploration of molecular libraries using molecular descriptors. Small mole-
cules are given specific numerical values called molecular descriptors that could reflect
simple physicochemical features or complex structural properties. These descriptors are
mathematical representations that depict the molecular structure and its properties [12].
Examples of molecular descriptors comprise surface area, molecular weight, atom distri-
butions, atom types, molecular weight, electro-negativities, bond distance, solvent proper-
ties, aromaticity indices, and many others [13]. The generation of molecular descriptors
involves experimentation, the use of quantum mechanical instruments, or prior informa-
tion [14, 15]. Molecular descriptors are classified based on the “dimensionality,” with 1D
(one-dimensional), 2D (two-dimensional), and 3D descriptors being the most common
types. 1D descriptors pertain to single scalar physicochemical properties of molecules, for
example, their log p values, molecular weight, and molar refractivity. However, 2D de-
scriptors are based on the molecular structure or composition and may include 2D fin-
gerprints or topological indices. While 3D descriptors, on the other hand, are derived
from the molecular conformation, encompassing properties such as 3D fingerprints, di-
pole moments, electrostatic potential, and the energy of the lowest unoccupied molecular
orbital or highest occupied molecular orbital. Table 12.1 provides examples of software
used to forecast the molecular descriptors of small molecules.
Fingerprints are the most commonly utilized similarity search techniques, repre-
senting molecular structure and characteristics through bit string depictions [12, 16].
Fingerprints encode numerous descriptors of molecule as predefined bit settings. Bi-
nary features may be used to record these descriptors (e.g., “does the molecule possess
features a, b, and c?”; yes/no) [16]. Instead, values of descriptors can be numerically
encrypted.
In order to identify structurally related particles or structural similarity-based
cluster collection, molecular fingerprint-based approaches sought to characterize the
particles in a form that would allow fast structural comparison. Compared to pharma-
12 Computational approaches for enzyme inhibition discovery 297
https://t.me/med1917

cophore QSAR or mapping models, these methods use significantly fewer computer
resources. In contrast to many other LB-CADD systems, they solely rely on chemical
structure and leave out molecules that are known to perform biological roles, making
this approach more qualitative [17]. Moreover, fingerprint-based algorithms take into
account the similarity of all molecule components rather than focusing only on those
that are m ost crucial for the task. This decreases the likelihood of overfitting and
needs smaller datasets. Yet, the influence of unnecessary features has an impact on
model performance; as a result, the typically contracted chemical regions are evalu-
ated [17]. In spite of this disadvantage, 2D fingerprints continue to remain a viable
alternative for similarity-based virtual screening (SBVS) [18]. Chemotype information
from early hit compounds is employed in 2D similarity search databases. The result-
ing tests are then utilized to conduct 2D fingerprint and 3D shape similarity searches
in order to identify distinct agonists. Molecular descriptors enable the rapid compari-
son of small molecule structure and/or physicochemical characteristics. The Tanimoto
coefficient (T) is a prevalent method for evaluating the similarity between two mole-
cules. A T value greater than 0.85 indicates a strong match; however, it does not nec-
essarily indicate that the two molecules are biologically comparable. Bero et al. [19]
introduced weighed Tanimoto coefficient, and they estimated the similarity score in
amphetamines (ATS) and nonamphetamine (NATS) drugs. The data used was acquired
from the 3D Exact Legendre database. This dataset comprises features extracted from
Table 12.1: Software used to predict molecular descriptors [11].
Software
name
Total numbers (types of predicted descriptors)
ADAPT > (physicochemical, topological, electronic, geometrical)
ADMET
predictor
> (functional group counts, constitutional, E-state, topological, Meylan flags, Moriguchi
descriptors, electronic properties, acid base ionization, molecular patterns, hydrogen
binding, D descriptors, hydrogen bonding, empirical estimates of quantum descriptors)
PreADMET > (topological, physicochemical, geometrical, constitutional, etc.)
CODESSA >, (charge-related, thermodynamical, geometrical, constitutional, topological,
semiemperical)
MOE > (physical properties, topological, structural keys, etc.)
DRAGON >, (D autocorrelations, topological, geometrical, constitutional, constitutional, RDF,
WHIM, GETAWAY, D binary and D frequency fingerprints, functional groups properties,
etc.)
MARVIN
Beans
> (geometrical, topological, fingerprints, physicochemical, etc.)
MOLGEN-
QSPR
> (geometrical, topological, constitutional, etc.)
298 Mohammad Ovais Dar et al.
https://t.me/med1917

a total of 7,212 3D molecular structures, encompassing both ATS and non-ATS drugs.
Specifically, it includes 3,610 structures of ATS drugs and 3,602 structures of non-ATS
drugs. For the purposes of th is study , the ATS drugs were employed as the training
set, while the non-ATS drugs were used as the testing set. The summary of similarity
score of datasets with their corresponding class of similarity is given in Table 12.2.
12.2.1.1 Workflow
Various phases employed in molecular similarity-based search include:
(a) Standard formatting: Molecules are initially read and transformed to standard
formats for subsequent processing. Some of the standard formats include PDB, SDF,
mol, and mol2. This stage is import ant to excluding problematic structures such as
polymers, incorrect valances, and free radicles [11].
(b) Filtering: Small molecule databases consist of many undesired problematic com-
pounds, such as inorganic and isotope atoms, charged carbon atoms, as well as desir-
able small molecules. When using molecular filters in similarity-based search, it is
essential to eliminate these undesired small molecules. The principle of drug-likeness
isawidelyusedapproachtofiltersmallmolecular libraries, which is assessed by
using the Lipinski et al. [20] rule of five (also recognized as the rule of five or RO5).
According to the rule, a molecule that is suitable as a drug candidate should not have
above one deviation from the following measures: (1) it should not contain more than
10 hydrogen bond acceptors (oxygen or nitrogen atoms), (2) it should not contain
greater than 5 hydrogen bond donors (oxygen or nitrogen atoms having one or more
hydrogen), (3) an octanol–water partition coefficient (log p) should not be greater
than 5, and (4) molecular weight should be less than 500 Da. If a molecule violates two
or greater than two of these rules, its absorption is typically reduced. Other generally
Table 12.2: Summary of dataset similarity scores and corresponding similarity classes.
S. no. Class of drugs with molecule ID Similarity score Class of similarity
. ATS (pk) Absolute
. ATS (pk) . Very high
. ATS (pk) . High
. NATS (NATS) . Medium
. ATS (pk) . Medium
. NATS (NATS) . Low
. ATS (pk) . Low
. ATS (pk) . Very low
. NATS (NATS) . None
#ATS, amphetamines; NATS, non-amphetamines.
12 Computational approaches for enzyme inhibition discovery 299
https://t.me/med1917

used screening filters comprise extended drug-like filters [21], fragment-like filter [22],
Egan filter [23], Veber filter [21], and so on.
(c) Removal of duplicates: Several protonation states and subsequent tautomers are
possible for a given molecule. The user is responsible for determining whether or not
such duplicates are crucial for study.
12.2.1.2 Applications
(a) Keiser et al. [24] utilized chemical similarities of small molecules to identify new
targets. They categorized 65,000 ligands into groups representing hundreds of phar-
macological targets and linked various receptors based on similarities of the ligands.
Based just on ligand similarity, the authors constructed minimum spanning trees, pre-
dicted, and verified new biological targets for ligands.
(b) As was already mentioned, molecular similarity metrics like the Tanimoto coeffi-
cient are useful tools for grouping and creating networks of related small molecules.
Tanimoto coefficients and other chemical similarity measurements are now utilized
to envisage binding to various biological targets as well as off-target effects and ad-
verse medication responses [11].
12.2.2 Pharmacophore modeling
The international union of pure and applied chemistry defines pharmacophore as
“The ensemble of steric and electronic features that is necessary to ensure the optimal
supramolecular interactions with a specific biological target structure and to trigger
(or to block) i ts biological response” [25]. Pharmacophore model explains how the
chemical properties of ligands are arranged in space to interact so that it interacts
with the target receptor [26]. Aromatic ring systems, positively charged ionizable
groups, hydrophobic regions, negatively charged ionizable groups, and hydrogen
bond donors and acceptors, are some of the chemical characteristics employed in
pharmacophore modeling [27]. When the structural knowledge about the therapeutic
target is not known, the preferred approach is ligand-based pharmacophore (LBP).
The objective of LBP is to discover the primary 3D pattern of characteristics that is
required for the majority of input ligands to interact with the receptor. However, this
job becomes more challenging when the sum of input ligands and their flexibilities
increase. Thus, conformational search becomes a vital and expensive stage in LBP.
Various programs such as Phase, HypoGen, RAPID, DISCO, MPHIL, and HipHop are
used to compute all likely conformations of the input molecules. Pharmacophore-
based virtual screening may be used to discover ligands with varied skeletons, how-
ever, the same spatial arrangement of important functional moieties that interact
300 Mohammad Ovais Dar et al.
https://t.me/med1917

with the target. The target binding site’s bioactive conformation of the molecules can
be included in the pharmacophore model. This model is frequently also used in mo-
lecular alignment step in QSAR studies [28]. LigandScount, PharmMapper, Cataly st,
GALAHAD, and PHASE are some of the commonly employed programs which permit
programmed (automatic) construction of pharmacophore model (Table 12.3). Spatial
restrictions in areas occupied by inactive molecules are also included in a good phar-
macophore model, and the model is frequently further optimized to make it less re-
strictive. The final model either makes all pharmacophoric properties that are not
constantly seen in active molecules optional or eliminates them [2]. To reduce the pos-
sibility of false-positive and also false-negative outcomes, the developed pharmaco-
phore model should have the highest level of sensitivity and specificity. It must also
be verified using a separate external test set [29].
Sanapali and coworkers [94] used pharmacophore modeling to identify matrix metal-
loproteinase (MMP-9) inhibitors using phase module of Schrodinger to build the phar-
macophore hypothesis. They conducted contour plot analysis to assess how the spatial
arrangement of structural or molecular characteristics in the chosen molecules,
which exhibit MMP-9 inhibitory activity, impacts their behavior. They mapped all the
active and inactive molecules onto a developed pharmacophore model. Figure 12.1
shows the pharmacophore mapping of active and inactive compound on to the phar-
macophore. The orange circ les in the figure correspond to the aromatic rings, blue
dots are donor pharmacophoric features while as green dots imply hydrophobic phar-
macophoric feature. It is evident from Figure 12.1 that the active molecules are align-
ing with the developed pharmacophore and the inactive molecules are not properly
aligned with the model. Therefore, the pharmacophore mapping contour diagrams vi-
sually represent the critical structural or chemical features that contribute to a mole-
cule’s biological activity.
Table 12.3: Some software used for ligand-based
pharmacophore modeling [11].
Name Source
Quasi Denovopharma
LigandScout Inteligand
MOE Chemical computing group
Phase Schrodinger
Unity Certara
12 Computational approaches for enzyme inhibition discovery 301
https://t.me/med1917

12.2.3 Quantitative structure–activity relationship (QSAR)
QSAR is used to measure the relationship between a given chemical or biological pro-
cess and the chemical structures of a group of compounds. The underlying hypothesis,
equivalent structural or physiochemical qualities, should result in equivalent activity
and is behind the QSAR approach [30, 31]. The goal of QSAR models is to form a link
between biological activity and the measurable physicochemical properties of a group
of molecules, specifically the properties of the substituents [32]. Broadly the enormous
range of these physicochemical properties may be classified into three common types
which include steric, electronic, and hydrophobic properties. Primarily, a collection of
chemical components or lead molecules that show the required biological activity of
interest are found. The physicochemical characteristics and the biological activity of
theactivemoleculesarecorrelatedquantitatively.Theactivechemicalsaresubse-
quently optimized using the created QSAR model to maximize the pertinent biological
activity. The intended activity of the anticipatedmoleculesisnextexperimentally
tested. Thus, the QSAR approach may be used as a guide to find chemical changes that
have increased activity. QSAR model is constructed with the primary purpose of pre-
dicting the activity of chemical compound (Figure 12.2). It uses mathematical and sta-
tistical relationships to quantitatively link the structural and chemical characteristics
of compounds to their observed activity or property, facilitating predictions for new
or untested compounds based on their chemical structure.
Figure 12.1: Mapping of active compounds inactive compounds onto the pharmacophore.
302 Mohammad Ovais Dar et al.
https://t.me/med1917

The overall QSAR approach depends upon a series of successive stages:
(1) Finding of ligands having experimentally measured values of the intended biolog-
ical activity; though they should have sufficient chemical variety to show exten-
sive range of activities, these ligands should ideally belong to a congeneric series.
(2) Determination and identification of molecular descriptors related to several phys-
iochemical and structural properties of molecules.
(3) Det ermine relationships between biological activity and molecular descri ptors
which describes the difference in activity in dataset.
(4) Analysis of predictive power and statistical stability of QSAR model.
(5) Use of statistical model to calculate the biological action of novel molecules.
These models, which are used in chemical, biological, and engineering sciences, are
basic regression or classification models. As an example, QSAR regression models link
a group of predictor variables (X) to the efficacy of the response factor (Y). Some col-
lections of QSAR models link the predictor variables to the definite value of the re-
sponse factor [33]. Molecular descriptors, or the chemical characteristics of molecules,
are linked to observed activity. The statistical analytical method used to develop the
model depends on the data type. For example, classification-based approaches are
used to process graded response data, whereas regression-based methods are used to
handle quantitative data [11]. The three primary statistical techniques employed in
linear QSAR to identify crucial molecular features for activity are:
(i) MLR (multivariable linear regression analysis)
(ii) PCA (principal component analysis)
(iii) PLS (partial least square analysis)
The most popular technique for creating a regression-based QSAR model is MLR,
which is one of the re gression-based approaches. A linear relationship among many
dependent variables (biological activity) and independent variables (descriptors) is as-
Figure 12.2: Outline of prediction of activity using QSAR model.
12 Computational approaches for enzyme inhibition discovery 303
https://t.me/med1917
Соседние файлы в папке Библиотека им академика М.И. Перельмана
