Title: QSAR aided design of potent c-Met inhibitors using molecular

docking, molecular dynamics simulation and binding free energy

calculation

Authors: Liyuan Guo, Yulu Yang, Jianbo Tong, Zelei Chang, Peng
Gao, Yuan Liu, Yakun Zhang, and Xiaoyu Xing

This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). The VoR will be published online
in Early View as soon as possible and may be different to this Accepted
Article as a result of editing. Readers should obtain the VoR from the
journal website shown below when it is published to ensure accuracy of
information. The authors are responsible for the content of this Accepted
Article.

To be cited as: Chem. Biodiversity 2024, e202400782

Link to VoR: https://doi.org/10.1002/cbdv.202400782

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Computational Simulation Study of Potential Inhibition of c-Met

Kinase Receptor by Phenoxy pyridine Derivatives: Based on QSAR,

Molecular Docking, Molecular Dynamics

Li-yuan Guo1,2, Yu-lu Yang1,2, Jian-bo Tong*1,2, Ze-lei Chang1,2, Peng-Gao1,2, Yuan
Liu1,2, Ya-kun Zhang1,2, Xiao-yu Xing1,2
1
College of Chemistry and Chemical Engineering, Shaanxi University of Science and

Accepted Manuscript
Technology, Xi'an 710021, China
2
Shaanxi Key Laboratory of Chemical Additives for Industry, Xi'an 710021, China
*Corresponding author: Jian-Bo Tong*1,2;
E-mail: [email protected]

Abstracts
The mesenchymal-epithelial transition factor (c-Met) is a tyrosine kinase receptor
protein, and excessive cell transformation can lead to cancer. Therefore, there is an
urgent need to develop novel receptor tyrosine kinase inhibitors by inhibiting the
activity of c-Met protein. In this study, 41 compounds are selected from the reported
literature, and the interactions between phenoxy pyridine derivatives and
tumor-associated proteins are systematically investigated using a series of
computer-assisted drug design (CADD) methods, aiming to predict potential c-Met
inhibitors with high activity. The Topomer CoMFA(q2=0.620, R2=0.837) and HQSAR
(q2=0.684, R2=0.877)models demonstrate a high level of robustness. Further internal
and external validation assessments show high applicability and accuracy. Based on
the results of the Topomer CoMFA model, structural fragments with higher
contribution values are identified and randomly combined using a fragment splice
technique, result in a total of 20 compounds with predicted activities higher than the
template molecules. Molecular docking results show that these compounds have good
interactions and van der Waals forces with the target proteins. The results of
molecular dynamics and ADMET predictions indicate that compounds Y4, Y5, and
Y14 have potential as c-Met inhibitors. Among them, compound Y14 exhibits

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
superior stability with a binding free energy of -165.18 KJ/mol. These studies provide
a reference for the future design and development of novel compounds with c-Met
inhibitory activity.

Accepted Manuscript
Keywords: QSAR, c-Met, molecular docking, molecular dynamics simulation

Introduction
Tumors are one of the most dangerous diseases for human health, affect millio
ns of lives each year. The quest for effective tumor therapies has become the
focus of current research, with receptor tyrosine kinases being primary targets.
[1, 2]
The treatment of tumors has become a significant research hotspot . The c-
mesenchymal-epithelial transition factor (c-Met) is a common tyrosine kinase re
ceptor whose signal pathway is closely associated with cellular reorganization,
proliferation, and differentiation[3, 4, 5, 6,7]
.
c-Met exhibits abnormally high expression, amplification or mutation in tumor tissues
such as lung cancer[8], colon cancer[9] and breast cancer. Therefore, by inhibiting the
[1, 2]
activity can lead to disruption of its physiological transduction pathway . Zhang et
al[10] have shown the efficacy of small molecule inhibitors for c-Met enzyme activity
inhibition in cancer therapy. K Mkhayar [9] uses computer simulations to devise four
new compounds based on dime zone derivatization. These compounds are found to be
stable when bind to c-Met and show potential in inhibiting human colon cancer,
demonstrating promising effects. Mkhayar K. et al[8] investigate the quantitative
relationship between the bioactivity of anti-lung cancer fine cells and the molecular
structure of a series of compounds. Through QSAR model prediction, molecular

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
docking, and ADMET property assessment, the study identifies newly designed
compounds that show potential as anti-lung cancer drugs. Chu, C et al [1] report that
derivatives of pyridazinone structures containing a 4-phenoxypyridine nucleus
demonstrate significant c-Met inhibitory activity when test in vitro against the human
lung adenocarcinoma cell line A549 and the human intestinal cancer cell line HT-29,
both of which exhibit high c-Met expression. Consequently, c-Met kinase has
emerged as an attractive target for cancer therapy.

Accepted Manuscript
Inhibitors with c-Met targets are currently store on the market, such as
"cabozantinib"[11] "crizotinib"[12] and "onuzumab"[13] are available on the market.
While these inhibitors have shown beneficial effects, they also have some drawbacks
[14-16]
. For instance, patients often do not experience a rapid and immediate positive
[17, 18]
response during treatment . Specifically, "cabozantinib"[11] is frequently
accompanied by side effects such as vomiting and diarrhea in clinical trials[19].
Therefore, the development and design of new compounds with potential
pharmacodynamic value are urgently needed to achieve adequate inhibition of c-Met
activity and to produce new anticancer drugs that are low-cost, low-toxicity, and
highly effective[19].
2. Materials and Methods
2.1 Data set
In this study, 41 c-Met small molecule reported in the literature are selected for
molecular docking, construction of 3D-QSAR models and molecular dynamics
simulations, and they show different levels of biological activities for inhibiting c-Met
interactions[20, 21]. The corresponding semi-inhibitory concentration (IC50) values are
converted to pIC50 (-logIC50) values, considering the distribution and structural
diversity of the bioactivity values. Ten representative compounds of the selected 41
compounds are randomly chosen as the test set (25%), and the remaining 31
compounds (75%) are used as the training set for model building (Figure 1), which
conforms to a normal distribution. The main role of the test set is to test the predictive
ability of the model. The structures and associated activities of the small molecule are
listed in Table S1. The structures and associated activities of the small molecule

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
compounds are listed in Table2.

Accepted Manuscript
Figure 1. Distribution of training and test sets

2.2 Structural preparation

The 3D structures of the compounds are drawn by the Sketch Tool module of the
SYBYL 2.0 software package and the energy minimum is done for each compound
structure, the potential energy minimum of the compounds is achieved by the Powell
conjugate gradient minimum algorithm. The energy convergence condition is 0.005
kcal/mol Å. The atomic charges are calculated by the Gasteiger-Huckel method, and
the maximum number of iterations is set to 1000, and the stable conformations of the
compounds are obtained from the above parameters, and the default values of SYBYL
2.0 are used for the other parameters.
2.3 Topomer CoMFA
Topomer CoMFA (Translocator Comparative Molecular Force Field Method) is a
variant or extension of the traditional CoMFA (Comparative Molecular Field Analysis)
method, a novel 3D-QSAR method that can predict the biological activity of
[22]
compounds . Firstly, the common skeleton of the compounds is selected, after
which the parts outside the common skeleton are cut. Secondly, the potential energy
parameters of the steric and electrostatic fields of all the segments are obtained by

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
running the software. Finally, a quantitative constitutive model is developed to show
the relationship between molecular structure and biological activity, using the obtain
parameters as independent variables and biological activity as the dependent variable.
Compared to conventional CoMFA, Topomer CoMFA enhances its ability to predict
and capture important structural insights by considering multiple conformations and
exploring various binding modes, and improves the accuracy of predicting molecular
bioactivity.

Accepted Manuscript
2.4 HQSAR
Hologram QSAR (HQSAR) is a quantitative constitutive relationship (QSAR) model
based on chemical structure and molecular descriptors, a 2D-QSAR approach that
features the structure of a compound with the aim of predicting its biological activity.
Unlike traditional QSAR models, HQSAR is a new technique that employs specialist
fragment fingerprints (molecular holograms) as predictive variables of biological
activity. Since no pairing is required, HQSAR models are faster than other models and
HQSAR is more widely used[23, 24]. Calculations using this approach allow for fast,
accurate results, and predictive for many data sets.
2.5 Partial least squares analysis (PLS)
Partial Least Squares (PLS) is a method of building a model by explaining the
relationship between the Y variable and other variables. A simple explanation is given
to show that the method is a simple and sensible way of forming predictive
equations[25]. Biological activity (pIC50) is used as the dependent variable, and
descriptors generated by Topomer CoMFA and HQSAR are used as independent
variables. PLS is commonly used to assess the linear correlation between the
descriptors generated through the model and the biological activity values. The
cross-validation analyses are carried out using the leave-one-out (LOO) method with

cross-validation correlation coefficients ( q 2 )and the optimal number of components

( N ).The N values are then used in the non-cross validation analysis to calculate the

non-cross validation correlation coefficients ( r 2 ), standard error estimates ( SEE ) and

Fisher's test ( F ) values, and to calculate the energy contribution of each field to

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
[26]
better evaluate the accuracy and robustness of the constructed model . Afterwards,

the following equations (1)、(2) are used to calculate q2 and r2 of the constructed

model, respectively.

 ( yˆ − y ) … (1)
 2

q = 1− i =1
2 i

 (y − y)
 2

i =1 i

 ( y − y )( yˆ − yˆ ) 
2

= 
2 i i i
r … (2)
 ( y − y )   ( yˆ − yˆ )
2 2

Accepted Manuscript
i i i

yˆi and yi are the predicted activity values and experimental activity values for the

training set; yi and ŷ are the average activity values for experimental and predicted

activity values for the training set.

2.6 QSAR model validation
The ideal QSAR model not only has good internal validation results, high q2 and r2
values, but external validation is also important for evaluating the predictive ability of
the model. In this paper, the predictive power of the QSAR model is verified by
calculating the biological activity of compounds that are not used in the modeling
process. The r2pred is used as a commonly used external test parameter to evaluate the
predictive ability of the model, which can be calculated by Equation (3).

 SD − PRESS 
2
rpred =  … (3)
 SD 

SD expresses the sum of the squares of the deviations between the biological activity
of the molecules in the test set and the average activity of the molecules in the training
set, and PRESS is the sum of the squares of the deviations between the predicted
2
and experimental activities of the molecules in the test set. When rpred   , the

developed QSAR model has good predictive ability.

2.7 QSAR model has good predictive ability.
Roy also proposed a new, rigorous and robust statistical metric based on the
Golbraikh-Tropsha method. In addition, the validation is carried out by drawing on
the calculation criteria of these two methods, calculating according to equations

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
(4)-(12).

k=
 ( y  yˆ ) … (4)
i i

 ( yˆ )
2
i

k' =
 ( y  yˆ ) … (5)
i i

( y )
2
i

 ( y − k  yˆ )
2

R 2
= 1− i i
… (6)
( y − y )
0 2
i i

Accepted Manuscript
 ( yˆ − k  y ) ' 2

= 1−
'2 i i
R … (7)
 ( yˆ − yˆ )
0 2
i

rm2 = R 2 1 − ( R 2 − R02 … (8) )

rm2 = rm2 − rm'2 … (9)

 ( y − yˆ )
n 2

RMSE = i =1 i i
… (10)
n


n
yi − yˆi
MAE = i =1
… (11)
n

2 i =1 ( yi − y )( yˆi − yˆ )
n

CCC = … (12)
 ( y − y ) +  ( yˆ − yˆ ) + n ( y − yˆ )
n 2 n 2 2
i =1 i i =1 i

k and k ' denote the slope of regression of experimental activity against predicted

activity or the slope of regression of predicted activity against experimental activity

through the origin. R and R0 denote the regression correlation coefficients between

the experimental and predicted activity values of the test set.

2.8 Appliance taxonomy
Application domain evaluation is an important indicator of whether a model can
accurately predict the properties or activities of new compounds. Currently, there are
several approaches on the evaluation of the domain of application of models[27, 28].
The AD applicability domain is a theoretical region in chemical space, defined by
model descriptors and modelling responses, and therefore by the properties of the
compounds in the training set, as represented by specific molecular descriptors in

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
each model. On this basis this study selects application domains to validate the
application of this mode[29], its defining formula:

hi = xiT ( X T X ) xi … (13)
−1

Where hi is the leverage value and the standardized molecular descriptor vector for

the compound i , and X is the entire application domain matrix.

By comparing with the threshold ( h ), if h  h , it indicates that the predictability of

Accepted Manuscript
the model is stable in this region. On the contrary, the predictive ability of this model
is worse[30, 31]. The definition equation for the threshold:

h* = 3 p ' / n … (14)

In the formula, p' is the molecular descriptor add 1; n is the number of training set.

2.9 Molecular screening and activity evaluation

Virtual screening (VS) is a technique that uses computational methods to predict and
screen potential active compounds from a large library of compounds[32]. It improves
the efficiency of the drug discovery process by calculating the interactions and
properties of molecules for the purpose of quickly identifying potential biological
active compounds. Firstly, a reliable Topomer CoMFA model is established in this
study by SYBYL 2.0 software package. Secondly, through Topomer CoMFA
similarity matching, molecular fragments with high activity contribution and high
similarity are searched for in the Zinc20 database using the R group as a template.
Finally, the molecular fragments are then screened and combined using the Topomer
CoMFA Search module in the Sybyl 2.0 software package in a bid to theoretically
obtain new compounds with high inhibitory activity.
2.10 Molecular docking
Molecular docking is the process of simulating interactions between two molecules or
parts of molecules in a computer to predict the bind mode and affinity between them
[33, 34]
. There are three most widely used docking methods, Autodock 4[35], Autodock
Vina[36], Surflex-dock[37] The protein used for docking(3LQ8, resolution: 2.02 μm, no
R-value: 0.252, working R-value: 0.191, observed R-value: 0.194), is selected from

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
the PDB database.
Preparation of ligands
All docking protocols employ flexible docking to accommodate the flexibility of the
ligand, while maintaining rigid docking conditions for the receptor. Autodock 4
[19, 34]
requires several preparatory steps before performing molecular docking . Pymol
is used to minimize the energy of prepared proteins and small molecules, including
the addition of water molecules and hydrogen atoms, under the Kollman united atom

Accepted Manuscript
force field[38]. The unbuilt fit cut-off value is set at 8.0, and the dielectric constant is
set at 1.0. The basic Powell method is applied for energy minimization of proteins,
with 10,000 iterations performed to achieve this. Using AutoDockTools-1.5.7, polar
hydrogen atoms are added to the receptor backbone and Gasteiger charges are added
to the 3LQ8 receptor to achieve proper system equilibrium and to check the
protonation state of the amino acids as well as the ligand. A 3D grid defined as
X1/460, Y1/440, and Z1/440 was established using the AUTOGRID algorithm of
AutoDockTools-1.5.7[39] and the docking sites of the selected small molecules in the
target proteins are determined according to the coordinates of the small molecule in
the target protein to the action site(X1/4 -1.382 nm, Y1/4 4.24 nm, Z1/4 27.008
nm).The reliability of the ligand-protein docking is assessed using root mean square
deviation (RMSD). Finally, Discovery Studio Visualizer 2017 software is employed to
visualize the hydrogen bonding and hydrophobic interactions between the ligand and
the active site of the target protein in graphical form.
2.11 ADMET forecast
ADMET (Absorption, Distribution, Metabolism, Excretion and Toxicity) prediction
refers to the computational and modelling prediction of the absorption, distribution,
metabolism, excretion and toxicity properties of a compound in an organism. By
using computational methods for ADMET prediction, researchers can reduce their
reliance on laboratory testing, evaluate ADMET properties of lots of compounds more
rapidly, predict and evaluate compounds in bulk, and reduce development costs and
failure rates. Therefore, this study predicts the ADMET properties of newly designed
compounds by using the online web servers admet SAR[40] and ADMET Lab[41].

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2.12 Molecular dynamics simulation
Molecular dynamics (MD) simulations can generate large amounts of data. To
extrapolate meaningful conclusions from the simulations, MD trajectories need to be
analyzed against the individual positions (and possibly velocities and forces) of all
atoms, or a selected subset of atoms, within each time frame of the trajectory[42].
In order to further investigate the interactions of the ligand molecules with the target
protein complexes and the stability of the complexes in a natural dynamic

Accepted Manuscript
environment, this study uses molecular dynamics (MD) simulations to study the
complexes[43]. Molecular dynamics simulations are performed in Gromacs20.7[44],
where the topology file and force field parameters of the protein are constructed using
[45]
the pdb2gmx command in the CHARMM36 (Jul-2021) force-field protein
parameter set, and the force-field parameters of the ligand are determined by the
CHARMM total force-field (C Gen FF) [46] parameter set, with the time-step set to 2fs.
The system is immersed in a periodic boundary condition (PBC), a 12-sided box
consisting of a Tip3p water model, and then sodium and chloride ions are added to
make the system electrically neutral. The steepest descent minimum algorithm is used
to minimum the energy of the system in 50000 steps, so that the system is in a steady
state. The system is then equilibrated with 100 ps each of NVT and NPT to maintain a
temperature of 300 K and a pressure of 1 atm. Finally, 100 ns simulations are
performed for each selected complex at constant temperature and pressure.
Then, this study analyses the topology files and trajectory files generated during the
simulation using Gromacs20.7 built-in programs. These analyses include root mean
square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration
(Rg), solvent accessible surface area (SASA), number of hydrogen bonds, binding
energy (MM/PBSA), and gibbs free energy landscape (FEL) analysis.
3. Results and discussion
3.1 Topomer CoMFA results
The results of the Topomer CoMFA model are related to the cutting method of the
molecule, and different cutting methods have a great impact on the predictive ability
and its results. The contribution values reflect the effect of different fragments on the

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
inhibitory activity. The most active compound 21 is cut as a template molecule and
the rest are cut according to the common backbone of the template molecule. Table 1
shows the results of two different cut methods. The statistical parameters for model 1

are q 2 =0.620, r 2 =0.837, F =33.289, SEE =0.483 and N =4; those for model 2 are

q 2 =0.600, r 2 =0.839, F =33.856, SEE =0.479 and N =6. Based on the analysis of the

results of the model cuts, the results of model 1 are better, Figure 2a shows a linear

Accepted Manuscript
regression plot of the experimental activity values of the compounds against the
predicted values, with the distribution of the data points for the great majority of the
compounds around the regression line. Figure 2b shows a scatter plot of experimental
activity values versus residual values, with data points distribute on both sides of the
zero line. Individual offsets farther away, but within manageable limits. The results
show that the Topomer CoMFA model obtained by cutting model 1 has good
predictive ability.
Table 1. Calculated molecular cutting modes and parameters of Topomer CoMFA
model

NO. Cutting mode SEE F q2 r2

1 0.483 33.289 0.620 0.837

2 0.479 33.856 0.600 0.839

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 2a. experimental activity versus predicted activity; Figure 2b residuals of
experimental activity versus predicted activity for Topomer CoMFA model 1
3.2 HQSAR modelling
The HQSAR model is generated by using two feature parameters, fragmentation
discrimination and atomic size. These two characteristic parameters are directly
related to pIC50. At the same time, these parameters are to be developed and selected.
In this study, the default FL (4-7) is initially used to obtain models with higher
predictive power through different combinations of HL and FD (atom, bond, chiral,
linkage, hydrogen bond donor and hydrogen bond acceptor)[47, 48]. 48 HQSAR models
are built by different combination forms as shown in Table S3, where the best

resultant parameters are q 2 =0.684, r 2 =0.877, SEE =0.436, HL =151 and N =6. Keep

the existing model parameters constant, different segment lengths are adjusted to get
better model results. Table S4 demonstrates the effect of 11 sets of different fragment
lengths on the predictive ability of the model. The fragment lengths of the best model
(4) combination are obtained as 4-7, and all other resultant parameters are

q 2 =0.684, r 2 =0.877, SEE =0.436, HL =151 and N =6. Figure 3a shows the linear

regression relationship between experimental and predicted activities of the HQSAR

model with a good linear fit. Figure 3b shows a scatter plot of experimental activity
values versus residual values, with essentially no systematic error in the residual
range on either side of the zero line.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 3a. experimental activity versus predicted activity; Figure 3b. residuals of experimental
activity versus predicted activity for HQSAR model 1

3.3 Validation of the QSAR model

This experiment preliminarily demonstrates that the QSAR model have good internal
prediction ability and stability, based on which the external validation of the model is
carried out. Table 2 shows the results of external validation. The external validation
parameter rpred2 for the Tompomer CoMFA and HQSAR models is 0.802 and 0.884,
respectively, which initially indicates that the models have effective predictive ability.
The R2 are 0.837 (Topomer CoMFA), 0.877 (HQSAR), and the slope of the regression
curve. The k and k', have values of 0.967 and 1.029 (Topomer CoMFA), 0.988 and
1.009 (HQSAR). The parameters in Table 2 show robust results. In addition, the root
means square error ( RMSE ), minimum mean square error ( MAE ) and consistency
correlation coefficients ( CCC ) of the model are 0.510, 0.068 and 0.910 (Topomer
CoMFA) and 0.323, 0.262 and 0.961 (HQSAR). Therefore, the two models internal
and external validation parameters can further demonstrate the stability and predictive
ability of the QSAR model in this study, which can be used for further activity
prediction and result analysis.
Table2. Statistical parameters of QSAR model validation methods

Parameters Criterion Topomer CoMFA HQSAR

Statistics from the best model q 2   0.620 0.684

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
r 2   0.837 0.877
SEE 0.483 0.436
N 4 6
2
External validation rpred 0.802 0.884

Golbraikh-Tropsha method R 2  0.6 0.850 0.942

R0 2 0.849 0.915

Accepted Manuscript
R0 '2 0.833 0.934

(R 2
− R02 ) (R 2
− R02 )
  0.001 0.029
R2 R2

(R 2
− R0'2 ) (R 2
− R0'2 )
  0.019 0.009
R2 R2

k 0.85  k  1.15 0.966 0.988

k' 0.85  k '  1.15 1.029 1.009

rm2 rm2   0.787 0.813

rm2 rm2   0.043 0.067

RMSE Close to 0 0.510 0.323

MAE 0.068 0.262
CCC CCC   0.910 0.961
3.4 Test for Y Randomization
The test for Y randomization is accomplished by randomly arranging the y data,
holding all x data constant, and performing the entire model-building process,
iteratively breaking the link between the response variable y and the independent
variable x (QSPR/QSAR: the molecular descriptor), just as is done for the actual y
data. Thus, y-randomization can be considered a special case of the permutation test.
Estimates of R2 and q2 are generated for each run and record. If R2 and q2 are given
lower in each data than the original data, then it means that the QSAR model has
some reliability[49, 50]. As shown in Table S5, the runs produce data that are all lower
than the original data. Thus, these results further confirm the stability of the QSAR

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
model.
3.5 Application domain validation
With the lever method (Figure 4), it is possible to verify whether the compounds
activities are located within or outside the domain of the structural model. Indeed,
leverage is used as a quantitative measure in the domain of model applicability and is
suitable for assessing the extent of extrapolation[30]. Figure 4a shows a linear
regression plot of leverage values normalized versus residual values for Topomer

Accepted Manuscript
CoMFA, with most compounds falling within the hatched vertical line. It is shown
that the established QSAR model has a relatively wide range of application areas.
However, one point (compound 38, compound 35) is found to be outside the critical
hat value in the Topomer CoMFA and HQSAR plots, respectively. The contribution
value can be specifically for this specific structure. Therefore, test set and training set
predictions for these compounds can be considered equally reliable.

Figure 4. (a: Topomer CoMFA, b: HQSAR) Linear regression plot of leverage values
against normalized residual values
3.6 Tomper CoMFA model contour map
In Tomper CoMFA model, two types of contour plots are generated based on the
spatial and steric field characteristics of the training set, and the template molecule 21
(pIC50 = 7.21) with the highest compound activity is chosen as the reference, as is
discussed under separate headings below.
Space field contour map

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The spatial profile of the template molecule 21[20, 21] obtained from the literature is
shown in Figure 5. Green area contours indicate favorable positions for larger group
substitution and yellow area contours indicate unfavorable positions for larger group
substitution. Figure 5a shows the stereo field of the R1 fragment in compound 21,
with the green region of carbon atom 3, indicate that the larger the group at this
position, the stronger the inhibitory activity will be. And the substitution of the methyl
group (compounds 10-18) in the R1 group, which is attached to the nitrogen atom, by

Accepted Manuscript
the ethyl group (compounds 19-27), results in a significant increase in the value of the
activity. Such as compound 27 (pIC50 = 6.23) > compound 18 (pIC50 = 5.72),
compound 26 (pIC50 = 5.90) > compound 17 (pIC50 = 5.48). Figure 5c shows the
steric field of the R2 fragment in compound 21, the aromatic compound C ring is
surrounded by a yellow region, the yellow color indicates that a smaller group in the
corresponding region enhances the inhibitory activity, and the activity is reduced
when the C ring is replaced by a substituent (compounds 13,14,15,16,17,18). For
example, compound 10 (pIC50 = 6.80) > compound 11 (pIC50 = 6.31), compound 19
(pIC50 = 7.02) > compound 20 (pIC50 = 6.82).
Electrostatic field contour plot
The electrostatic field contour plot shows that there are negatively charged (red area)
and positively charged (blue area) substitutive regions in the 3D space where the
inhibitor activities are located, which indicates they are in the electrostatic field
[51]
contour plot . Figure 5b represents the electrostatic field of the R1 fragment in
compound 21, and the presence of red regions near the nitrogen atom, carbon atoms 1
and 3 indicates that the introduction of negatively charged groups in the
corresponding regions is more favorable for the activity. For example, compound 34

(pIC50 = 5.53)＞compound 38 (pIC50 = 3.02). The presence of a blue region near

carbon atom 2 indicates that positively charged groups are beneficial for increasing

activity. For example, compound 19 (pIC50 = 7.02) ＞compound 10 (pIC50 = 6.80).

Figure 5d represents the electrostatic field of the R2 fragment in compound 21, with
the presence of red regions near the hydroxyl group of the A ring and near the

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
nitrogen 7, 9, and 8 carbons of the B ring, as well as the carbon atom 3 of the C ring,
which indicates that the introduction of electron-withdrawing groups in the
corresponding regions is more favorable for activity. The presence of blue regions
near the B-8 and 9 carbon atoms and near the C-2 carbon atom and C-4 fluorine
suggest that electron-donating groups in this region are more favorable for activity.

Accepted Manuscript
Figure 5. The 3D contour maps of the steric and electrostatic fields of the Topomer CoMFA

models with compound 21 as the template.

3.7 HQSAR model atomic contribution map analysis

The HQSAR model is represented by two feature parameters, fragment differentiation
and atomic size generation, which are correlated with pIC50. Compound 21 with the
highest activity (pIC50 = 7.21) and compound 38 with the lowest activity (pIC50 = 3.02)
are selected to analyze the relationship between activity and atomic distribution.
Green or yellow color in the atomic contribution map indicates favorable contribution,
orange and red color indicates unfavorable contribution, and white color indicates no
special contribution[52]. The B ring in Figure 6a shows yellow and green colors,
which indicates a positive effect on the compound to increase its activity. The

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
aromatic substituents on the A ring in Figure 6b are shown in red, which indicates
that this region exhibits unfavorable contributions. These are consistent with the
results of the Topomer CoMFA contour analysis. The nitrogen atom itself is
electronegative, so the introduction of nitrogen atoms into a compound plays a
positive role for the inhibitor.

Accepted Manuscript
Figure 6. represent the atomic contributions of compound 21 and compound 38.
3.8 Molecular design and activity prediction
Through the established Topomer CoMFA model, this study hopes to design novel
anti-tumour drugs with promising applications. Referring to the fragment contribution
descriptor values calculated by the Topomer CoMFA model, representative fragments
with high contribution values are selected to be combined with fragments with high
contribution values searched from the ZINC20 databases to constitute new
compounds with high expected inhibitor activity. Figure 7, 10 R1 fragments with high
contribution values are screened and combined with 2 R2 fragments with high
contribution values selected from the training set in the hope of obtaining new highly
active compounds. Table S6 lists the structures and predicted activities of the 20 new
compounds obtained in different combinations, and 14 of the newly design
compounds have higher predicted inhibitory activities than the template molecule,
which further confirms the reliability of the model. Thus, the newly designed
compounds can move on to the next step in the research. In addition, these

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
compounds require further synthetic studies combined with in vitro and ex vivo
bioactivity evaluations to determine their inhibitory effects.

Accepted Manuscript
Figure 7. Combined design of new compound molecules of R1, R2
3.9 Molecular docking
Molecular docking is used to assess whether stable complexes can be formed between
proteins and small-molecule ligands, and to predict the binding conformation of
small-molecule ligands at target protein binding sites[53,54]. In this study, Surflex-dock,
Autodock 4, and Autodock Vina are used to evaluate and check the reliability of
docking. The target protein structures are obtained from the PDB database (RCSB)
(https://www.rcsb.org/structure/3LQ8,resolution:2.02 μm, no R-value: 0.252, working
R-value: 0.191, observed R-value: 0.194). The small molecules and target proteins are
re-docked using each of the three methods, and the small molecule ligands before and
after docking are superimposed, and the changes in the conformation of the ligand
small molecules after docking are observed, using the value of the RMSD between the
ligand poses as a reference. The root mean square deviation (RMSD) value is a
criterion for determining the suitability of the docking pattern between the original

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ligand and the redocked ligand[55]. The parameters of the three docking schemes are
shown in Table 3. When RMSD < 2, the docking method is reliable. After docking
with Autodock 4.2, the RMSD value is 0.509. As shown in Figure. 8 for Autodock 4
docking, the original ligand almost completely coincides with the docked ligand,
which indicates that the conformations docked by this method are very close to the
bioactive conformations of the original ligand, and this docking result provides some
reference for the selection of docking methods. The docking scores and amino acid

Accepted Manuscript
interactions of the newly designed compounds, as calculated by Autodock 4 and
Autodock Vina, are presented in Table 3. Autodock 4 is selected as the docking
method for this study because it produces more hydrogen bonds and dehydration
interactions, which are crucial for inhibitory activity. Additionally, it provides a
valuable reference for subsequent molecular dynamics simulations.
Table 3. RMSD values (Å) and docking parameters between each re-docking position and the

initial position in the different docking protocols.

Autodock 4.2 Autodock Vina Surflex-dock

Crysta
Ligand RMSD Kcal/mol RMSD Kcal/mol RMSD Kcal/mol
lstructure
3LQ8 88Z 0.506 -10.51 0.457 -11.93 1.260 12.475

Figure 8. Superimposed conformation of redocked ligand and primitive ligand at protein target

(PDB: 3LQ8), redocked ligand (yellow) and primitive ligand (pink)

Figure 9, demonstrates the interaction of the template molecule 21 with the active site

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
of the target protein, it is not difficult to find that the inhibitor interacts with amino
acids Lys1110 (Lys=H...O-H, 1.82 Å, 161.17°), Phe1223 (Phe=O...H- O, 2.69 Å,
103.86°) , Asp1222 (Asp=O...N-H, 3.32 Å, 94.71°), and Met1160 (Met=O...H-O, 2.95
Å, 136.16°) to produce hydrogen bonding interactions. The binding energy fraction of
the docking is -9.55 Kcal/mol.

a b

Accepted Manuscript
Figure 9. Binding conformation and mutual sway of template molecule 21 within the active site of

protein (PDB:3LQ8) (a) 3D view (b) 2D view

The bind scores and interactions result from the docking of the newly design
compounds by Autodock 4.2 are displayed in the table S6, and the bind energies of
the compounds are scored in the range of -10.15 Kcal/mol to -11.59 Kcal/mol. It
suggests that the compounds designed in this study interact well with the target
protein binding site and have a similar common backbone with the template molecule,
lower binding energy and may have similar inhibitory potential. Y04, Y05, Y06, Y10
and Y14, which have relatively good predicted activity and docking results, are
selected for binding to the active site sites of target proteins.
Figure 10. compound Y04 is the compound with predicted activity and good docking
scoring, which forms hydrophobic interactions with amino acid residues Met1131,
Leu1157, Val1155, Ala1108, Val1092, Met1211, Ile1084 and Tyr341, and forms a
hydrogen bond with the active site residues Asp1222(Asp=N1 ...O-H, 4.19 Å, 110.84°;
Asp=N2...O-H, 4.19 Å, 90.03°).

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
a b

Accepted Manuscript
Figure 10. Binding conformation and mutual sway of compound Y04 within the active site of the

protein (PDB:3LQ8) (a) 3D view (b) 2D view

Figure 11. compound Y14 forms hydrophobic interactions with amino acid residues
Leu1157, Val1092, Val1108, Met1211, and Ile1084, and forms three hydrogen bonds
hydrophobic with active site residues Lys1110 (Lys=H1...O-H, 4.89 Å, 112.83°;
Lys=H3 ...O-H, 4.95 Å, 140.81°), Asp1222 (Asp=N...O-H, 2.04 Å, 144.31°), and
Tyr1159 (Tyr=C...O-H, 1.71 Å, 107.69°).

a b

Figure 11. Binding conformation and mutual sway of compound Y14 within the
active site of the protein (PDB:3LQ8) (a) 3D view (b) 2D view
Figure 12. Compound Y10 forms hydrophobic interactions with amino acid residues
Met1211, Val1092, Ala1108, Ala1221, Leu1140, Leu1157 and Ile1084, and forms
three hydrogen bonds with active site residues Asp1222 (Asp=N...H-C, 2.98 Å, at
128.39°; Asp=O...N-H, 1.74 Å, 8.35°), Lys1110 (Lys=H...O-H, 1.64 Å, 154.04°),
Met1160 (Asp=N...O-H, 2.63 Å, 164.44 °).

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 12. Binding conformation and mutual sway of compound Y10 within the active site of the

protein (PDB:3LQ8) (a) 3D view (b) 2D view

3.10 Molecular dynamics modelling

With the aim of further investigating the results of molecular docking, the molecules
with better docking results: Y04, Y05, Y06, Y14, Y10 and homologous ligands bound
to the target proteins into complexes are selected for molecular dynamics simulations.
After performing 100 ns of molecular dynamics simulations, the trajectories and
coordinate files are analyzed using the built-in program of Gromacs, including root
mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of
gyration (Rg), solvent-accessible surface area (SASA), the number of hydrogen
bonding, and the Gibbs free energy, with the intention of exploring several aspects of
the ligand-complexes, such as flexibility and stability.
3.10.1 Root mean square deviation
RMSD is used to assess the positional change of the ligand-complex between the
conformation and the initial conformation during the simulation[56]. Figure13a shows
the RMSD of the protein backbone during the simulation. The compounds all
maintain low values during the simulations, showing relative steady states and
flexible bind sites. Compounds Y05 and Y06 remain relatively smooth throughout the
simulation, and compound Y10 fluctuates up and down within 0.15 ns after 50 ns.
Especially, compound Y14 has a lower RMSD value compared to the cognate ligand
T(88z), with RMSD Y04 < 0.25 and RMSD T < 0.3 after 65 ns. It indicates that the
ligand molecules converge to a state of stable binding during simulation within the

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
active site. Figure13b illustrates the RMSD values of the protein ligands during the
simulation, and all the newly designed molecular complexes show good stability
compared to the template molecules.

Accepted Manuscript
Figure 13. a is RMSD values of small molecule ligands extracted from 100 ns simulations of

protein-ligand complexes; b is RMSD values of protein backbones extracted from 100 ns

simulations of protein-ligand complexes.

3.10.2 Root-mean-square fluctuation

RMSF is the change in position of protein residues during a reaction molecular
dynamics simulation. Small RMSFs indicate that the corresponding amino acid
residues or atoms are relatively stable during the simulation. Figure 14a shows the
RMSF plots of the newly designed compounds with a similar distribution of values,
which indicates that the newly designed compounds and homologous ligand can
equally affect the fluctuation of bound protein residues. 1050-1110, 1150-1225 are the
key positions of the active pocket of the receptor protein. In this region amino acid
residues Met1160, Asp1222, Lys1110 form hydrogen bonds with the ligand molecule
and amino acid residues Ile1084, Ala1108, Ala1221, Ala1222, Leu1157 form
hydrophobic bonds with the ligand molecules, which is the reason for the low
fluctuation in this fraction

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 14a. RMSF of ligand-protein complexes at 100 ns; Figure 14b is the radius of gyration

3.10.3 Radius of gyration

Rg is an assessment of the stability and denseness of the target protein conformation
after ligand binding to the target protein. In general, spherical proteins or compact
proteins have smaller Rg values, looser proteins have larger Rg values[57]. In Figure
14b, it is apparent that the Rg values of the newly design compounds basically tend to
be smooth. And the Rg values are distributed in the range of 1.9-2.1 nm. Y04 and Y10
show small fluctuations between 30-50 ns (fluctuations between 0.1 ns), respectively.
Compared to the homologous ligand T(88Z), Y05, Y06, and Y14 show a relatively
stable level of fluctuation of the radius of gyration values in the range of less than
0.05 nm. Thus, the ligand-protein complexes are somewhat stable and dense, which
contributes to the stability of molecular dynamics simulations.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 15a. the number of hydrogen bonds in the ligand-protein complex at 100ns, Figure 15b

shows the solvent accessible surface area

3.10.4 Solvent accessible surface area (SASA)

Solvent accessible surface area (SASA) is the surface area where complex biological
macromolecules can interact with solvents, and the hydrophobicity of amino acid
residues plays an important role in protein folding. An increase in SASA indicates the
structure of the complex expands relative to the contact area of the solvent molecules,
and vice versa, a tightening of the structure leading to a decrease in the surface area of
the solvent in contact. Figure 15b, the SASA values of the protein-ligand complexes
basically level off and are distributed between 149-161 nm2, which indicates that the
newly designed compounds can bind stably to the receptor proteins.
3.10.5 Hydrogen bond analysis
Hydrogen bond is an important factor in maintaining the conformational stability of
protein ligands, and the stability of ligand-target protein binding can likewise be
judged by the number of hydrogen bonds. Figure 15a shows the number of hydrogen
bonds for the five complexes and the template molecule. The hydrogen bond range of
the complexes are homologous ligand (0-4), Y04 (0-5), Y05 (0-5), Y05 (0-5), Y14
(0-7), Y10 (0-6), and the number of hydrogen bonds of complex Y14 shows good
predominance and is higher than the number of homologous ligands. These data all
demonstrate that five newly design ligands bind to proteins with stable interactions.
3.10.6 MM/PBSA binding free energy calculation.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
The binding free energy is an important indicator for evaluating the stability of
ligand-protein binding. The last 10ns MD traces are extracted for binding free energy
analysis, and Table 4 provides a summary of the free energy values for the ligand
complexes, target proteins, and template molecules, with negative values indicating
favorable binding interactions and positive values indicating unfavorable binding
interactions. The binding energy of Y14 is close to that of the homologous ligand, and
Y05 has a lower binding energy compared to the homologous ligand and the template

Accepted Manuscript
molecule. Therefore, it can be shown that the newly designed ligand complexes bind
to the protein with high efficiency.
Table 4. Binding energies of newly designed compounds (Y04, Y05, Y06, Y14, Y10) for

MM/PBS studies with protein (3LQ8), template molecule M

NO. EVDM Eele E polar Enpolar EGbind

88Z -223.562 -72.597 216.981 -28.027 -107.865

M -225.325 -75.197 231.802 -32.553 -101.273
Y04 -253.464 -123.015 293.965 -34.121 -116.635
Y05 -285.451 -61.057 218.108 -36.78 -165.18
Y06 -256.627 -89.985 240.695 -34.319 -140.237
Y14 -253.464 -123.015 293.965 -34.121 -116.635

Y10 -234.395 -93.273 214.259 -35.017 -148.426

* Each
unit of energy in the diagram is KJ/mol.

3.10.7 Free energy landscape analysis

Free energy landscape (FEL) maps are analyzed for RMSD and radius of gyration (Rg)
of protein C backbone atoms as global minima. The energy bars indicate the Gibbs
free energy (kcal/mol) from the lowest energy (blue) to the highest energy (red)
conformational states, and the color blocks represent the flexibility of conformations
at different energy states. If multiple lowest energy clusters are present, the
protein-ligand interaction is unstable; if a single lowest energy cluster is present, the
protein-ligand interaction is stable. Figure 16 shows the free energy distribution of
the complexes with the lowest energy conformation extracted. The Gibbs free energy

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
distribution of complex Y04 is in the range of 0-11.6 kcal/mol, with RMSD
fluctuating in the range of 0.00-0.437 nm and Rg fluctuating in the range of 1.92-2.07
nm; The Gibbs free energy distribution of complex Y14 is in the range of 0-12.1
kcal/mol, with RMSD fluctuating in the range of 0.00-0.379 nm and Rg fluctuating in
the range of 1.95-2.05 nm; The Gibbs free energy distribution of complex Y10 ranges
from 0-12.9 kcal/mol, the RMSD fluctuates in the range of 0.00-0.477 nm, and the Rg
fluctuates in the range of 1.95-2.11 nm. Overall, the complexes all present a single

Accepted Manuscript
and advantageous energy cluster. The results again indicate that the newly designed
ligands bind well to the protein sites.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Accepted Manuscript
Figure 16. a, b, c, d show the free energy landscape (FEL) and lowest energy conformation of the

newly designed compounds Y04, Y05, Y14, Y10

3.11 In silico ADMET prediction

The ADMET predictions for the newly designed compounds are shown in Table S8
and include the assessment of certain parameters such as blood-brain barrier (BBB)
permeation, absorption level (HIA), water solubility, CYP3A4 binding and plasma
protein binding[58]. In terms of absorption parameters, the compounds show good
human intestinal absorption (HIA) (68.56%-98.73%) except for Y05 which show
moderate level of absorption compared to other compounds. All compounds show
good permeability in Caco2 cell model. Having moderate blood-brain barrier (BBB)
diffusion levels suggests that most compounds are expected to have minimal CNS
side effects. The plasma protein binding model shows that compound Y05 binds
plasma protein values <90%. On the other hand, the remaining compounds are
expected to bind more than 90% to plasma proteins. For the newly designed
compounds have the potential to be substrates for CYP2D6 and CYP3Y4. At lower Cl
values, the drugs are more stable in vivo and all the compounds show lower total

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
clearance indices. In addition, the toxicity predictions of the compounds are evaluated,
and most of the compounds pass the Ames toxicity test with fewer side effects in
humans, while individual compounds may need to be tested for toxicity in vivo and in
vitro.
Conclusion
In the present work, the molecular structures of bioactive and phenoxy pyridine
derivatives are systematically investigated use 3D-QSAR, molecular docking, MD,

Accepted Manuscript
ADMET and binding free energy. The modelling results, including the contour plots
from the Tomoper CoMFA (q2=0.620, r2=0.837, N=4 and SEE=0.483) model and
atomic contribution plots from the HQSAR model (q2=0.684, r2=0.877, N=6 and

SEE=0.436)，are utilized to identify the key groups or atoms inhibiting the bioactivity.

Based on the Topomer CoMFA model, 20 new compounds with potential inhibitory
effects are designed, all of which exhibit predicted inhibitory activities higher than the
template molecules. The molecular docking results indicate that the newly designed
compounds mainly form hydrophobic interactions and van der Waals force
interactions with amino acid residues Asp1222, Leu1157, Ala1108, Val1092, Lys1110,
etc., and these hydrophobic and hydrogen bonds play a key role in the inhibitory
activity. Molecular dynamics simulations show that the newly design compounds can
bind well to proteins with excellent stability and further verify the reliability of the
molecular docking results. ADMET properties shows its potential medicinal value for
c-Met inhibition. Therefore, the results of this study provide valuable insights and
information for the development of c-Met protein inhibitors and the utilization of
phenoxy pyridine derivatives.

Disclosure statement
The authors declare that they have no conflict of interest.

Author Contribution Statement

Writing the original draft: YYL. Material preparation and data collection: GP, LY.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Writing review and editing: ZYK, CZL and XXY. Funding acquisition: TJB, GLY.

Funding
This work was supported by the National Natural Science Foundation of China
(22373062) and the Graduate Innovation Fund of Shaanxi University of Science and
Technology.

Accepted Manuscript

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
References
1. Chu, C.; Rao, Z.; Pan, Q.; Zhu, W., An updated patent review of small-molecule c-Met
kinase inhibitors (2018-present). Expert Opinion on Therapeutic Patents 2022, 32 (3), 279-298.
2. El Aissouq, A.; Chedadi, O.; Kasmi, R.; Elmchichi, L.; En-nahli, F.; Goudzal, A.;
Bouachrine, M.; Ouammou, A.; Khalil, F., Molecular Modeling Studies of C-Glycosylfavone
Derivatives as GSK-3β Inhibitors Based on QSAR and Docking Analysis. Journal of Solution
Chemistry 2021, 50 (5), 808-822.
3. Christensen, J. G.; Burrows, J.; Salgia, R. J. C. l., c-Met as a target for human cancer and
characterization of inhibitors for therapeutic intervention. 2005, 225 (1), 1-26.

Accepted Manuscript
4. D’Angelo, N. D.; Bellon, S. F.; Booker, S. K.; Cheng, Y.; Coxon, A.; Dominguez, C.;
Fellows, I.; Hoffman, D.; Hungate, R.; Kaplan-Lefko, P. J. J. o. m. c., Design, synthesis, and
biological evaluation of potent c-Met inhibitors. 2008, 51 (18), 5766-5779.
5. Doa’a, G. A.-.; Al-Husein, B. A.; Qasaimeh, G. R. J. B.; Pharmacotherapy, A mini-review of
c-Met as a potential therapeutic target in melanoma. 2017, 88, 194-202.
6. Ko, B.; He, T.; Gadgeel, S.; Halmos, B. J. A. o. t. m., MET/HGF pathway activation as a
paradigm of resistance to targeted therapies. 2017, 5 (1).
7. gomha, s.; El-Sayed, A.-A.; Zaki, M.; Alrehaily, A.; Elbadawy, H.; Al-Shahri, A.;
Alsenani, S.; Abouzied, A., Synthesis, In vitro and In silico Studies of Novel bis ‐
triazolopyridopyrimidines from Curcumin Analogues as Potential Aromatase Agents. Chemistry &
Biodiversity 2024.
8. Mkhayar, K.; Daoui, O.; Elkhattabi, S.; Chtita, S.; Elkhalabi, R. J. B. R. A. C., QSAR study
and Molecular Docking assisted design of novel cyclohexane-1, 3-dione Derivatives as anticancer
agents for non-small cell lung cancer. 2023, 13, 524.
9. Mkhayar, K.; Elkhattabi, K.; Elkhalabi, R.; Haloui, R.; Daoui, O.; Edache, E. I.; Chtita,
S.; Elkhattabi, S. J. C. J. o. A. C., Evaluation of dimedone-derived compounds as inhibitors against
human colon cancer: Insights from 2D-QSAR, ADMET prediction, Osiris, Molinspiration, and
molecular modeling. 2023, 51 (11), 100330.
10. Zhang, Q.-W.; Ye, Z.-D.; Shi, L., c-Met kinase inhibitors: an update patent review
(2014-2017). Expert Opinion on Therapeutic Patents 2019, 29 (1), 25-41.
11. Shang, R.; Song, X.; Wang, P.; Zhou, Y.; Lu, X.; Wang, J.; Xu, M.; Chen, X.;
Utpatel, K.; Che, L. J. G., Cabozantinib-based combination therapy for the treatment of
hepatocellular carcinoma. 2021, 70 (9), 1746-1757.
12. Ayoub, N. M.; Al-Shami, K. M.; Alqudah, M. A.; Mhaidat, N. M. J. O.; therapy, Crizotinib,
a MET inhibitor, inhibits growth, migration, and invasion of breast cancer cells in vitro and
synergizes with chemotherapeutic agents. 2017, 4869-4883.
13. Shah, M. A.; Cho, J.-Y.; Tan, I. B.; Tebbutt, N. C.; Yen, C.-J.; Kang, A.; Shames, D. S.;
Bu, L.; Kang, Y.-K. J. T. o., A randomized phase II study of FOLFOX with or without the MET
inhibitor onartuzumab in advanced adenocarcinoma of the stomach and gastroesophageal
junction. 2016, 21 (9), 1085-1090.
14. Chu, C.; Rao, Z.; Pan, Q.; Zhu, W. J. E. o. o. t. p., An updated patent review of
small-molecule c-Met kinase inhibitors (2018-present). 2022, 32 (3), 279-298.
15. Liang, C.; Wang, H.; Zhang, M.; Cheng, W.; Li, Z.; Nie, J.; Liu, G.; Lian, D.; Xie, Z.;
Huang, L.; Zeng, X., Self-controlled release of Oxaliplatin prodrug from d-α-tocopheryl

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
polyethylene glycol 1000 succinate (TPGS) functionalized mesoporous silica nanoparticles for
cancer therapy. Journal of Colloid and Interface Science 2018, 525, 1-10.
16. Zhao, C.-h.; Zhang, C.-l.; Shi, J.-j.; Hou, X.-y.; Feng, B.; Zhao, L.-x., Design, synthesis,
and biofunctional evaluation of novel pentacyclic triterpenes bearing
O-[4-(1-piperazinyl)-4-oxo-butyryl moiety as antiproliferative agents. Bioorganic & Medicinal
Chemistry Letters 2015, 25 (20), 4500-4504.
17. Liang, C.; Wang, H.; Zhang, M.; Cheng, W.; Li, Z.; Nie, J.; Liu, G.; Lian, D.; Xie, Z.;
Huang, L. J. J. o. c.; science, i., Self-controlled release of Oxaliplatin prodrug from d-α-tocopheryl
polyethylene glycol 1000 succinate (TPGS) functionalized mesoporous silica nanoparticles for
cancer therapy. 2018, 525, 1-10.

Accepted Manuscript
18. Zhao, M.; Zheng, Y.-H.; Zhao, Q.-Y.; Zheng, W.; Yang, J.-H.; Pei, H.-Y.; Liu, L.; Liu,
K.-J.; Xue, L.-L.; Deng, D.-X. J. E. j. o. m. c., Synthesis and evaluation of new compounds bearing
3-(4-aminopiperidin-1-yl) methyl magnolol scaffold as anticancer agents for the treatment of
non-small cell lung cancer via targeting autophagy. 2021, 209, 112922.
19. Stitou, M.; Toufik, H.; Bouachrine, M.; Lamchouri, F., Quantitative structure–activity
relationships analysis, homology modeling, docking and molecular dynamics studies of
triterpenoid saponins as Kirsten rat sarcoma inhibitors. Journal of Biomolecular Structure and
Dynamics 2021, 39 (1), 152-170.
20. Duan, Y.; Xu, S.; Xiong, H.; Wang, L.; Zhao, B.; Wang, P.; Wang, C.; Peng, Y.; Cai,
S.; Luo, R. J. B.; Letters, M. C., Discovery of novel 2-substituted-4-phenoxypyridine derivatives
as potential antitumor agents. 2018, 28 (3), 254-259.
21. Xiong, H.; Zhang, J.; Zhang, Q.; Duan, Y.; Zhang, H.; Zheng, P.; Tang, Q. J. B.;
Letters, M. C., Design, synthesis and biological evaluation of 4-(pyridin-4-yloxy) benzamide
derivatives bearing a 5-methylpyridazin-3 (2H)-one fragment. 2020, 30 (9), 127076.
22. Cramer, R. D., Topomer CoMFA: a design methodology for rapid lead optimization. Journal
of medicinal chemistry 2003, 46 (3), 374-88.
23. Lowis, D. R. J. T. T. N., HQSAR: a new, highly predictive QSAR technique. 1997, 1 (5), 17.
24. Sainy, J.; Sharma, R. J. S.; Research, Q. i. E., QSAR analysis of thiolactone derivatives using
HQSAR, CoMFA and CoMSIA. 2015, 26 (10), 873-892.
25. Garthwaite, P. H., An Interpretation of Partial Least Squares. Journal of the American
Statistical Association 1994, 89 (425), 122-127.
26. Waller, C. L., A Comparative QSAR Study Using CoMFA, HQSAR, and FRED/SKEYS Paradigms
for Estrogen Receptor Binding Affinities of Structurally Diverse Compounds. Journal of Chemical
Information and Computer Sciences 2004, 44 (2), 758-765.
27. Tropsha, A. J. M. i., Best practices for QSAR model development, validation, and exploitation.
2010, 29 (6‐7), 476-488.
28. Jaworska, J.; Nikolova-Jeliazkova, N.; Aldenberg, T. J. A. t. l. a., QSAR applicability domain
estimation by projection of the training set in descriptor space: a review. 2005, 33 (5), 445-459.
29. Netzeva, T. I.; Worth, A. P.; Aldenberg, T.; Benigni, R.; Cronin, M. T.; Gramatica, P.;
Jaworska, J. S.; Kahn, S.; Klopman, G.; Marchant, C. A. J. A. t. L. A., Current status of methods
for defining the applicability domain of (quantitative) structure-activity relationships: The report
and recommendations of ecvam workshop 52. 2005, 33 (2), 155-173.
30. Gramatica, P. J. Q.; science, c., Principles of QSAR models validation: internal and external.
2007, 26 (5), 694-701.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
31. Gramatica, P.; Giani, E.; Papa, E. J. J. o. M. G.; Modelling, Statistical external validation and
consensus modeling: a QSPR case study for Koc prediction. 2007, 25 (6), 755-766.
32. Tong, J.-B.; Liu, Y.; Xiao, X.-c.; Gao, P.; Xu, H.-y. J. M. S., Exploration of anti-tumour
inhibitors from colchicine derivatives based on 3D-QSAR, molecular docking and molecular
dynamics simulations. 2023, 49 (17), 1647-1665.
33. Ton, A. T.; Gentile, F.; Hsing, M.; Ban, F.; Cherkasov, A. J. M. i., Rapid identification of
potential inhibitors of SARS‐CoV‐2 main protease by deep docking of 1.3 billion compounds.
2020, 39 (8), 2000028.
34. Chu, H.; He, Q.-x.; Wang, J.-w.; Deng, Y.-t.; Wang, J.; Hu, Y.; Wang, Y.-q.; Lin, Z.-h.,
3D-QSAR, molecular docking, and molecular dynamics simulation of a novel

Accepted Manuscript
thieno[3,4-d]pyrimidine inhibitor targeting human immunodeficiency virus type 1 reverse
transcriptase. Journal of Biomolecular Structure and Dynamics 2020, 38 (15), 4567-4578.
35. Morris, G. M.; Huey, R.; Lindstrom, W.; Sanner, M. F.; Belew, R. K.; Goodsell, D. S.;
Olson, A. J. J. J. o. c. c., AutoDock4 and AutoDockTools4: Automated docking with selective
receptor flexibility. 2009, 30 (16), 2785-2791.
36. Peng, X. X.; Feng, K. R.; Ren, Y. J. J. R. a., Molecular modeling studies of quinazolinone
derivatives as novel PI3Kδ selective inhibitors. 2017, 7 (89), 56344-56358.
37. Irwin, J. J.; Sterling, T.; Mysinger, M. M.; Bolstad, E. S.; Coleman, R. G. J. J. o. c. i.;
modeling, ZINC: a free tool to discover chemistry for biology. 2012, 52 (7), 1757-1768.
38. Rani, A.; Aslam, M.; Khan, J.; Pandey, G.; Singh, P.; Maharia, R. S.; Pant, B. N.,
Computational Insights into Chromene/pyran Derivatives: Molecular Docking, ADMET Studies,
DFT Calculations, and MD Simulations as Promising Candidates for Parkinson's Disease.
Chemistry & Biodiversity 2024.
39. Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; Belew, R. K.; Olson,
A. J. J. J. o. c. c., Automated docking using a Lamarckian genetic algorithm and an empirical
binding free energy function. 1998, 19 (14), 1639-1662.
40. Yang, H.; Lou, C.; Sun, L.; Li, J.; Cai, Y.; Wang, Z.; Li, W.; Liu, G.; Tang, Y. J. B.,
admetSAR 2.0: web-service for prediction and optimization of chemical ADMET properties. 2019,
35 (6), 1067-1069.
41. Dong, J.; Wang, N.-N.; Yao, Z.-J.; Zhang, L.; Cheng, Y.; Ouyang, D.; Lu, A.-P.; Cao,
D.-S. J. J. o. c., ADMETlab: a platform for systematic ADMET evaluation based on a
comprehensively collected ADMET database. 2018, 10, 1-11.
42. Mathada, B. S.; Basha, N. J.; Javeed, M.; Karunakar, P.; Venkatesulu, A.; Erappa, K.;
Varsha, A. J. J. o. B. S.; Dynamics, Novel pyrimidines as COX-2 selective inhibitors: synthesis, DFT
analysis, molecular docking and dynamic simulation studies. 2024, 42 (4), 1751-1764.
43. Michaud‐Agrawal, N.; Denning, E. J.; Woolf, T. B.; Beckstein, O. J. J. o. c. c., MDAnalysis: a
toolkit for the analysis of molecular dynamics simulations. 2011, 32 (10), 2319-2327.
44. Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. J. J.
o. c. c., GROMACS: fast, flexible, and free. 2005, 26 (16), 1701-1718.
45. Huang, J.; MacKerell Jr, A. D. J. J. o. c. c., CHARMM36 all‐atom additive protein force field:
Validation based on comparison to NMR data. 2013, 34 (25), 2135-2145.
46. Hu, F.; Liu, X.-T.; Zhang, J.-L.; Zheng, Q.-C.; Eglitis, R. I.; Zhang, H.-X. J. J. o. C. I.;
Modeling, MD simulation investigation on the binding process of smoke-derived germination
stimulants to its receptor. 2019, 59 (4), 1554-1562.

This article is protected by copyright. All rights reserved.

Chemistry & Biodiversity 10.1002/cbdv.202400782

16121880, ja, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cbdv.202400782 by Lund University, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
47. Daoui, O.; Nour, H.; Abchir, O.; Elkhattabi, S.; Bakhouch, M.; Chtita, S. J. J. o. B. S.;
Dynamics, A computer-aided drug design approach to explore novel type II inhibitors of c-Met
receptor tyrosine kinase for cancer therapy: QSAR, molecular docking, ADMET and molecular
dynamics simulations. 2023, 41 (16), 7768-7785.
48. Daoui, O.; Elkhattabi, S.; Bakhouch, M.; Belaidi, S.; Bhandare, R. R.; Shaik, A. B.; Mali,
S. N.; Chtita, S. J. A. o., Cyclohexane-1, 3-dione derivatives as future therapeutic agents for
NSCLC: QSAR modeling, in silico ADME-Tox properties, and structure-based drug designing
approach. 2023, 8 (4), 4294-4319.
49. Rücker, C.; Rücker, G.; Meringer, M. J. J. o. c. i.; modeling, y-Randomization and its
variants in QSPR/QSAR. 2007, 47 (6), 2345-2357.

Accepted Manuscript
50. Rücker, C.; Rücker, G.; Meringer, M. J. J. C. I. M., Y-randomization–a useful tool in QSAR
validation, or folklore. 2007, 47, 2345-2357.
51. Yang, Y.; Qin, J.; Liu, H.; Yao, X. J. J. o. c. i.; modeling, Molecular dynamics simulation,
free energy calculation and structure-based 3D-QSAR studies of B-RAF kinase inhibitors. 2011,
51 (3), 680-692.
52. Chavda, J.; Bhatt, H. J. S. C., 3D-QSAR (CoMFA, CoMSIA, HQSAR and topomer CoMFA), MD
simulations and molecular docking studies on purinylpyridine derivatives as B-Raf inhibitors for
the treatment of melanoma cancer. 2019, 30 (6), 2093-2107.
53. Morris, G. M.; Lim-Wilby, M. J. M. m. o. p., Molecular docking. 2008, 365-382.
54. El Aissouq, A.; Chedadi, O.; Bouachrine, M.; Ouammou, A.; Khalil, F., Development of
novel monoamine oxidase B (MAO-B) inhibitors by combined application of docking-based
alignment, 3D-QSAR, ADMET prediction, molecular dynamics simulation, and MM_GBSA binding
free energy. Journal of Biomolecular Structure and Dynamics 2023, 41 (10), 4667-4680.
55. Ayaz, M.; Alam, A.; Zainab, Z.; Elhenawy, A. A.; Rehman, N. U.; Rahman, S. U.; Ali,
M.; Latif, A.; Harrasi, A. A.; Ahmad, M., Designing and Synthesis of Novel Fexofenadine‐
Derived Hydrazone‐Schiff bases as Potential Urease Inhibitors: In‐Vitro, Molecular Docking and
DFT Investigations. Chemistry & Biodiversity 2024.
56. Tong, J. B.; Gao, P.; Xiao, X. C.; Liu, Y.; Xu, H. Y. J. C., Molecular modeling studies of
Indoline Scaffold derivatives as PD‐1/PD‐L1 pathway inhibitors by QSAR, molecular docking and
molecular dynamics simulation techniques. 2024, 9 (8), e202302813.
57. Ouassaf, M.; Belaidi, S.; Chtita, S.; Lanez, T.; Abul Qais, F.; Md Amiruddin, H. J. J. o. B.
S.; Dynamics, Combined molecular docking and dynamics simulations studies of natural
compounds as potent inhibitors against SARS-CoV-2 main protease. 2022, 40 (21),
11264-11273.
58. Alanazi, M. M.; Eissa, I. H.; Alsaif, N. A.; Obaidullah, A. J.; Alanazi, W. A.; Alasmari, A.
F.; Albassam, H.; Elkady, H.; Elwan, A. J. J. o. E. I.; Chemistry, M., Design, synthesis, docking,
ADMET studies, and anticancer evaluation of new 3-methylquinoxaline derivatives as VEGFR-2
inhibitors and apoptosis inducers. 2021, 36 (1), 1760-1782.

This article is protected by copyright. All rights reserved.