Quantitative activity-activity relationship (QAAR) driven design to develop hydroxamate derivatives of pentanoic acids as selective HDAC8 inhibitors: synthesis, biological evaluation and binding mode of interaction studies

Article abstract of DOI:10.1039/d1nj02636d

Histone deacetylase 8 (HDAC8) has been implicated as a potential drug target of many diseases including cancer. HDAC8 isoform selectivity over other class-I HDACs is a major concern nowadays. In this work, a series of pentanoic acid based hydroxamates wit

Full text of DOI:10.1039/d1nj02636d

NJC
View Article Online
View Journal
PAPER
Quantitative activity–activity relationship (QAAR)
driven design to develop hydroxamate derivatives
of pentanoic acids as selective HDAC8 inhibitors:
synthesis, biological evaluation and binding mode
of interaction studies†
Cite this: DOI: 10.1039/d1nj02636d
a
a
Sk. Abdul Amin,
Prakruti Trivedi,b Nilanjan Adhikari,
Ganesh Routholla,b
Dhanya Vijayasarathi,b Sanjib Das,a Balaram Ghosh *b and Tarun Jha
*
a
Histone deacetylase 8 (HDAC8) has been implicated as a potential drug target of many diseases
including cancer. HDAC8 isoform selectivity over other class-I HDACs is a major concern nowadays. In
this work, a series of pentanoic acid based hydroxamates with different substituted cap groups have
been designed, synthesized, characterized, and screened against class-I HDACs. A quantitative activity–
activity relationship (QAAR) model was developed to design HDAC8 selective inhibitors. The designed
compounds obtained through the molecular modeling study were synthesized, characterized, and their
enzymatic as well as cytotoxic activities were measured. Two compounds 7i and 7f are found to be
selective HDAC8 inhibitors over other class-I HDACs. These compounds possess better antiproliferative
activities against some cancer cell lines. These observations are in agreement with the molecular
docking studies for the binding mode of interactions. Further studies show that compounds 7i and 7f
induce significant cell growth arrest in the G2/M phase, indicating their anticancer potentials. In
summary, our study confirms pentanoic acid based hydroxamate as selective HDAC8 inhibitors and two
compounds 7i and 7f may serve as lead molecules for further investigation.
Received 29th May 2021,
Accepted 12th August 2021
DOI: 10.1039/d1nj02636d
rsc.li/njc
on histone/nonhistone proteins along with histone acetyltrans-
ferase (HAT).3
Introduction
Epigenetic regulation of genes via post-translational modulation
The aberrant expression of various HDACs has been
in proteins results in a heritable change in gene function that observed in many human diseases, especially in cancer, which
culminates in a phenotypic change without altering DNA makes them important therapeutic targets for many human
sequences. Epigenetics plays important roles in cellular and cancers.4,5 Overexpression of classical HDACs has been
molecular regulatory processes.1–3 It is associated with DNA represented in the pathological situation. Hence, histone
methylation, histone modification, chromatin structural deacetylase inhibitors (HDACIs) have emerged as promising
remodeling, and noncoding RNA regulation. Among these, cancer therapeutic agents.6
histone acetylation and deacetylation are widely occurring
Zinc dependent HDACs are divided into three main classes:
post-translational modification processes. Histone deacetylases class-I HDACs (HDAC1, 2, 3, 8), class-II consists of class-IIa
(HDACs) constitute about 23% of the total domain of epigenetic HDACs (HDAC4, 5, 7, 9) and class-IIb HDACs (HDAC6, 10),
targets after methyltransferases.2 These are mainly responsible class-IV HDAC (HDAC11).7 The class-III HDACs are called
for balancing the acetylation/deacetylation of lysine amino acids sirtuins and depend on nicotinamide adenine dinucleotide
(NAD). These have different mechanism from classical zinc
dependent HDACs.7 HDAC8, a class I HDAC enzyme, is
expressed in various tissues and involved in the genesis of
malignancy. HDAC8 is localized in the nucleus and expressed
ubiquitously (cytogenetic location: Xq 13.1). It also alters
epigenetic modification.6 HDAC8 pivotally controls cellular
processes, such as muscle contraction, energy homeostasis,
and separation of sister chromatid.3,5 It is one of the essentials
a Natural Science Laboratory, Division of Medicinal and Pharmaceutical Chemistry,
Department of Pharmaceutical Technology, Jadavpur University, Kolkata 700032,
India. E-mail: tjupharm@yahoo.com
b Epigenetic Research Laboratory, Department of Pharmacy, BITS-Pilani,
Hyderabad Campus, Shamirpet, Hyderabad 500078, India.
E-mail: balaram@hyderabad.bits-pilani.ac.in
† Electronic supplementary information (ESI) available. See DOI: 10.1039/
d1nj02636d
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
Fig. 1 Structure of hydroxamic acid containing HDAC inhibitors, which are used in clinics or in clinical trials.
for cell survival, and it is implicated in the recently described a series of designed compounds having aliphatic side chain as
Cornelia de Lange syndrome (CdLS), which is a congenital a linker with a hanging aliphatic moiety between the ZBG and
malformation disorder.6 It plays an important role in some cap region, (iii) screening for HDAC inhibition potency and
diseases, including smooth muscle cell contractions, acute selectivity, (iv) cytotoxicity evaluation and understanding of the
myeloid leukemia, neuroblastoma, and T cell leukemia.8
Recently, HDAC8 is emerged as an anticancer target to two promising compounds, and (vi) possible binding mode of
design molecules against variety of haematological interactions of promising compounds in the drug-binding
structure–activity relationships (SARs), (v) cell cycle analysis of
a
malignancies.3 However, target specific HDAC8 inhibitor is still pockets of the HDAC8 enzyme. This was done to see whether
a challenge to medicinal chemists. Many of the HDAC inhibitors we are getting possible lead molecules or not against murine
currently in the clinics are broad-spectrum or pan-HDAC melanoma, non-small cell lung carcinoma (NSCLC), and acute
inhibitors (mainly hydroxamates), which inhibit many of the lymphoblastic leukemia (ALL). This work is a part of our quest
class I, II, and IV isoforms (Fig. 1).
of anticancer drug development4,8,13–16 and surely enriches the
Among these compounds, vorinostat/SAHA, panobinostat/ research community to find better active agents in the future.
LBH589, romidepsin/FK228, belinostat/PXD101, and pracinostat/
SB939 have been approved by the US-FDA, whereas chidamide/
CS055/HBI-8000 is approved by the China Food and Drug
Results and discussion
Administration (Fig. 1).3 These compounds have been shown to
The present investigation employs QAAR models to design the
inhibit most of the HDAC isoforms with nanomolar potency;
selective HDAC8 inhibitors that reliably correlated the
interestingly, the isoform that is inhibited least potently in each
case is the Class I isoform. However, most of these pan-HDACIs
biological data.
failed to pass the Phase II clinical trials due to unwanted adverse
effects. Despite these failures, studies on identifying effective and
selective HDAC8 inhibitors are still going on.
Design of molecules: quantitative activity–activity relationship
(QAAR) studies
Drug discovery is a time-consuming process, which requires
huge money and manpower.9,10 To meet this challenge,
quantitative structure–activity relationship (QSAR) methods for
correlating the physicochemical properties with desired target
endpoint are a good initiative.11,12 Herein, the application
of computationally inexpensive chemical descriptors in
combination with robust supervised statistical and machine
learning approaches was demonstrated for the development of
QSAR models.
To achieve the goal to find selective HDAC8 inhibitors, here,
we cover six aspects – (i) development of statistically robust
quantitative activity–activity relationship (QAAR) model to
design new compounds, (ii) synthesis and characterization of
In order to design selective HDAC8 inhibitors, regression
analysis based models were developed and validated. These
predictive and validated models were utilized to understand the
structural requirements of selective HDAC8 inhibitors and to
predict the activity of the designed molecules.
The dataset containing fifty-nine diverse molecules having
HDAC8 and HDAC3 inhibitory activities was collected from the
published work of the University of Illinois at Chicago17–26 to
perform the QAAR studies (Table S1 of the ESI†). The 2D
structures of investigated compounds were drawn by ChemDraw
Ultra 8.0 software27 followed by geometry optimizations. Then, the
2D descriptors were calculated by ‘‘PaDEL-descriptor’’.28 Finally,
multivariate regression analyses were performed to correlate the
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
Table 1 Contributions of descriptors of eqn (1)
Descriptor
Definition
Contribution
nAtomLAC
Number of atoms in the longest aliphatic chain
C(BC)(:C)(:N)
N–N
Z16 H; hierarchic element counts – these bits test for the presence or count
of individual chemical atoms represented by their atomic symbol.
Geary autocorrelation of lag 6 weighted by van der Waals volume
[#1]–N–C–[#1]
Geary autocorrelation of lag 3 weighted by I-state
Cc1ccc(C)cc1
Negative
Negative
Positive
Negative
PubchemFP357
PubchemFP300
PubchemFP2
GATS6v
PubchemFP528
GATS3s
Positive
Positive
Positive
Positive
PubchemFP713
independent descriptors, which in our case are the ‘‘PaDEL- model. R2 and Q2 values after several Y-randomization tests
HDAC8/HDAC3
descriptors’’ with the dependent variable (pIC
).29,30 for eqn (1) are shown in ESI† (Table S4).
50
The predictability of the developed model was evaluated using
both internal and external validation matrices.11,31–33
Euclidean distance-based applicability domain suggests that
all the training and test set compounds are within the boundary
Lastly, the statistically robust QAAR model was obtained of this hypothetical domain, as depicted in Fig. 2.
using a linear technique, which is mentioned below:
Hence, no outlier is found for this dataset.
The nAtomLAC is a 2D descriptor showing the negative
contribution to the HDAC8 selectivity. It defined the total
number of atoms in the longest aliphatic chain. The GATS6v,
a 2D autocorrelation descriptor, signifies the Geary autocorre-
lation of lag 6 weighted by van der Waals volume. It shows
the positive contribution to the selectivity, which indicates
that molecules with higher van der Waals volume will show
higher selectivity. Another 2D autocorrelation descriptor,
GATS3s defines Geary autocorrelation of lag 3 weighted by
I-state. This descriptor has a positive contribution to the HDAC8
selectivity over HDAC3. It carries information concerning the
topology of an atom and the electronic interactions due to all
other atoms in the molecule. The bond orders and bond
aromaticity are also found to be important. Lastly, to design
the HDAC8 selective inhibitors, these favorable fingerprints were
taken into consideration so that the specificity to HDAC8 may be
ensured.
(HDAC8/HDAC3)
pIC
= À4.471 (Æ1.966) À 0.435 (Æ 0.065) nAtomLAC
À 2.322 (Æ0.176) PubchemFP357 + 2.261 (Æ0.282)
50
PubchemFP300 À 4.118 (Æ0.441) PubchemFP2 + 7.587 (Æ1.716)
GATS6v + 1.724 (Æ0.301) PubchemFP528 + 2.831 (Æ0.888)
GATS3s + 0.797 (Æ0.313) PubchemFP713
(1)
2
NTrain = 45; R = 0.978; R2 = 0.956; RA = 0.946; SEE = 0.408;
F(8,36) = 96.720; p o 0.00; Q2 = 0.932; average rm(LOO)2 = 0.904;
2
2
2
Drm(LOO) = 0.030; NTest = 14; RTest = 0.971; RPred = 0.954;
2
2
average rm(Test) = 0.880; Drm(Test) = 0.028.
Eqn (1) corresponds to the best QAAR model achieved using
the S-MLR technique. The descriptors (Table 1) in this model
help to explore the structural requirements to attain HDAC8
selectivity over HDAC3.
The internal parameters, such as R2 = 0.956; Q2(LOO) = 0.932;
average r2
= 0.904; Dr2
= 0.030, indicate the better
m(LOO)
m(LOO)
Similarly, support vector machine (SVM) regression analysis
was performed on the training set (45 compounds) using the
same descriptors. Likewise, a search for the parameters tuning
was carried out in a stepwise manner. The SVM parameters
(The complexity parameter C = 1.0, exponent E = 1.0, epsilon
parameter = 0.001, tolerance T = 0.001) were used to construct
the SVM model. Here, we used John Platt’s Sequential Minimal
Optimization (SMO) algorithm with the Waikato Environment
for Knowledge Analysis (Weka) version 3.8.4.34,35 The predictive
performance derived from SVM as well as MLR models are
shown in Table 2.
predictive potential of the developed QAAR model. Correlation
matrix, t-values, and p-values of eqn (1) are shown in the ESI,†
Tables S2 and S3 respectively. The Y randomization results
(average R = 0.426; average R2 = 0.195; average Q2(LOO) = À0.397;
cR2 = 0.865) further confirm the robustness of the QAAR
p
Table 2 Predictive performance of MLR and SVM models using the
training and test data sets
Parameter
MLR
SVM
r
MAE
RMSE
r
MAE
RMSE
Train set
Test set
0.978
0.985
0.277
0.279
0.365
0.316
0.974
0.984
0.269
0.249
0.413
0.276
Fig. 2 Graphical representation of applicability domain of eqn (1) by the
r, correlation coefficient; MAE, mean absolute error; RMSE, root mean
Euclidean distance approach.
square error.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
of HDAC8 selective compounds that interact with both the
active site Zn2+ ion and this sub pocket. These unique features
can be exploited to design selective compounds for class I
HDACs.10 Now, it is clear that these compounds have short
length of linker moiety as well as side chains attached to the
linker improve the biological activity.11
From the results of QAAR studies, the understanding of SAR
of this series of compounds of the dataset becomes clearer and
specific conclusions can be made. Also, the reported QAAR
models pass all the requisite validation criteria. Hence, they are
useful for predicting the activity/selectivity of a similar class of
compounds.
Fig. 3 General structure of the designed inhibitors.
Results indicated that SVM regression analysis (r = 0.974,
The information obtained in the QAAR model was analyzed
MAE = 0.269 and RMSE = 0.413) performed similarly with the S- to design compounds, which were predicted to show the
MLR approach (r = 0.978, MAE = 0.277 and RMSE = 0.365). selectivity towards HDAC8. A series of compounds having an
Further evaluation of the predictive performance of the QAAR aliphatic side chain as a linker with a hanging aliphatic moiety
model on the training data set was performed using the same between the ZBG and cap region has been designed. Depending
data sampling approaches, including 5-fold cross-validation on our earlier observations3,4,8 as well as observation from
(S-MLR model: r5CV = 0.967; SVM model: r5CV = 0.964) and the current QAAR study, these compounds are designed. The
10-fold cross-validation (S-MLR model: r10CV = 0.958; SVM general structure of the designed inhibitors is depicted in
model: r10CV
=
0.951). Interestingly, all cross-validation Fig. 3.
Most of the designed compounds were predicted as ‘selec-
methods achieved good performances. Of the two multivariate
regression methods used in this study, SVM analysis was found tive’ HDAC8 inhibitors in these regression models, as shown in
to be similar to the traditional regression methods such as Table 3. The values of descriptors and the predicted activity
S-MLR. Hence, it may be postulated that the use of these ranges are given in the ESI† (Table S5).
molecular descriptors yield good predictive performances.
After the design of molecules, we turned our attention to
For this study, we concentrated on the 2D descriptor, validate the QAAR models experimentally. In this connection,
nAtomLAC. This descriptor suggests that the presence of more we synthesised the designed compounds, and subsequently,
number of atoms in the longest aliphatic chain contributes the purity of compounds was evaluated using 1H NMR, 13C
negatively to the HDAC8 selectivity over HDAC3. It has been NMR, and LC/MS analyses.
reported in the literature that class I enzymes have an internal
pocket at the bottom of the tube-like channel leading to the
Chemistry
zinc ion.4 This 14 Å hydrophobic internal cavity is formed A series of compounds having an aliphatic side chain as a
by the same residues for all class I isoforms with minor linker with a hanging aliphatic moiety between the ZBG and
differences. HDAC1, HDAC2, and HDAC3 in that order contain cap region have been synthesized (Scheme 1).
the longest cavities, and HDAC-8 has the smallest one.36 Recent
The purity of these final compounds was evaluated using
reports reveal that HDAC8 possesses a deep hydrophobic 1H NMR, 13C NMR, and LC/MS analyses. Spectral data of these
pocket adjacent to the active site channel allowing the design compounds are provided below:
Table 3 Designed molecules with their predicted HDAC8/HDAC3 selectivity
Predicted HDAC8/
HDAC3 selectivity
Predicted HDAC8/HDAC3
selectivity range
Average predicted
HDAC8/HDAC3 selectivityb MLR
Average Predicted HDAC8/HDAC3
selectivity range
Compa R1
R2
MLR
SVM
SVM
7a
7b
7c
7d
7e
7f
7g
7h
7i
2-Br
2-F
4-Br
4-Br
4-Cl
4-Cl
4-NO2
n-C5H11 1.035
i-C4H9 0.934
n-C5H11 0.912
n-C6H13 0.654
0.168
0.885
0.355
0.035
0.693
0.383
0.425
0.629
0.320
0.921
1.267
0.945
0.602
0.910
0.634
0.345
0.661
0.474
0.437
0.431
0.258
0.532
0.597
0.369
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
NS
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
n-C4H9
0.628
n-C5H11 0.564
n-C4H9
0.448
0.232
2,4-Dichloro n-C4H9
2,4-Dichloro n-C5H11 0.195
2,4-Dichloro C6H5 0.143
—
—
7j
7k
7l
n-C5H11 À0.074
n-C6H13 À0.208
NS
a
b
Compound number. Average predicted selectivity [(MLR + SVM)/2] as per the QAAR models; SEL, HDAC8 selective (positive value); NS, non-
selective (negative value).
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
Scheme 1 Reagents & conditions: (a) (i) thionyl chloride (ii) benzene (b) (i) L(+) glutamic acid (ii) 2 N NaOH (iii) 1 N HCl (c) (i) DCC (ii) Chloroform (d) (i)
R2NH2 (ii) 1 N Na2CO3 (iii) 1 N HCl (e) (i) thionyl chloride, (ii) anhydrous methanol, 12 h, 0 1C (f) (i) 50% w/v aqueous hydroxylamine solution, (ii) 1 N sodium
hydroxide, (iii) methanol (1 : 1), 2 h, RT (iv) 1 M hydrochloric acid.
2-(2-(2-Bromophenyl) acetamido)-N5-hydroxy-N1-pentylpen- 14.9, 6.9 Hz, 1H), 3.50–3.41 (m, 2H), 3.02 (m, 2H), 1.98–1.66 (m, 4H),
1
tanediamide (7a). Yield 56.09%, white solid, H NMR (400 MHz, 1.36 (m, 2H), 1.29–1.17 (m, 5H), 0.84 (t, J = 7.1 Hz, 3H), 13C NMR
DMSO-d6) d 10.42 (s, 1H), 8.25 (d, J = 8.1 Hz, 1H), 7.90 (t, J = 5.4 Hz, (101 MHz, DMSO-d6) d 171.26, 170.10, 168.93, 136.33, 131.78,
1H), 7.58 (d, J = 7.9 Hz, 1H), 7.32 (q, J = 7.5 Hz, 2H), 7.21–7.15 (m, 131.46, 119.96, 52.79, 41.71, 38.88, 29.32, 29.10, 28.94, 22.25,
1H), 4.22 (dd, J = 13.7, 8.0 Hz, 1H), 3.71–3.60 (m, 2H), 3.11–2.97 14.36. HPLC Rt: 2.474 LC/MS calculated for expected C18H26BrN3O4
(m, 2H), 2.06–1.67 (m, 4H), 1.43–1.16 (m, 7H), 0.85 (t, J = 6.9 Hz, [M]: 428.32; found: [M]+: 428, [M + 2]+: 430, [M + Na]+: 452; anal.
3H), 13C NMR (101 MHz, DMSO-d6) d 170.73, 168.77, 168.50, calcd for C18H26BrN3O4 (exact mass: 427.11; MW: 428.32): C, 50.47;
136.03, 132.18, 131.80, 128.54, 127.49, 124.45, 52.34, 42.10, 38.39, H, 6.12; N, 9.81; O, 14.94; found: C, 50.68; H, 6.08; N, 9.73; O,
28.86, 28.65, 28.47, 21.77, 13.88. HPLC Rt: 2.303, LC/MS calculated 14.67% (spectra of compound 7c are shown in the ESI,† Fig. S7–S9
for expected C18H26BrN3O4 [M]: 428.32; found: [M]+: 428, [M + 2]+: respectively).
430; anal. calcd for C18H26BrN3O4 (exact mass: 427.11; MW:
428.32): C, 50.47; H, 6.12; N, 9.81; O, 14.94; found: C, 50.56; H, diamide (7d). Yield 92.7%, cream powder, H NMR (400 MHz,
6.33; N, 9.70; O, 14.90% (1H NMR, 13C NMR, and mass spectra of DMSO-d6) d 10.41 (s, 1H), 8.29 (t, J = 8.5 Hz, 1H), 7.91 (t, J =
2-(2-(4-Bromophenyl)acetamido)-N1-hexyl-N5-hydroxypentane-
1
compound 7a are shown in the ESI,† Fig. S1–S3 respectively).
5.5 Hz, 1H), 7.47 (d, J = 8.4 Hz, 2H), 7.22 (d, J = 8.3 Hz, 2H), 4.17
N1-Butyl-2-(2-(2-fluorophenyl)acetamido)-N5-hydroxypentane- (dd, J = 14.0, 8.2 Hz, 1H), 3.46 (s, 2H), 3.09–2.95 (m, 2H), 2.20–
1
diamide (7b). Yield 57.46%, white powder, H NMR (400 MHz, 1.66 (m, 4H), 1.34 (dd, J = 13.2, 6.5 Hz, 2H), 1.23 (d, J = 4.6 Hz,
DMSO-d6) d 10.43 (s, 1H), 8.71 (s, 1H), 8.29 (d, J = 8.1 Hz, 1H), 6H), 0.85 (dd, J = 8.5, 5.0 Hz, 3H), 13C NMR (101 MHz, DMSO-d6)
7.92 (t, J = 5.8 Hz, 1H), 7.34–7.24 (m, 2H), 7.16–7.10 (m, 2H), 4.24 d 170.75, 169.57, 168.43, 135.84, 131.28, 130.95, 119.45, 52.30,
(td, J = 8.2, 5.7 Hz, 1H), 3.55 (s, 2H), 2.96–2.80 (m, 2H), 2.01–1.61 41.20, 38.40, 30.90, 28.90, 28.84, 28.45, 25.91, 22.00, 13.88. HPLC
(m, 5H), 0.82 (d, J = 6.7 Hz, 6H). 13C NMR (101 MHz, DMSO-d6) d Rt: 2.447 LC/MS calculated for expected C19H28BrN3O4 [M]:
170.93, 169.00, 168.49, 161.75, 159.32, 131.66, 128.52, 124.09, 442.35; Found: [M]+: 442, [M + 2]+: 444; anal. calcd for
115.01, 52.39, 45.91, 34.96, 28.84, 28.51, 27.96, 20.00 HPLC Rt: C19H28BrN3O4 (exact mass: 441.13; MW: 442.35): C, 51.59; H,
2.361 LC/MS calculated for [M]: 353.39; Found: [M + H]+: 354; 6.38; N, 9.50; O, 14.47; found: C, 51.78; H, 6.23; N, 9.79; O, 14.1%
anal. calcd for C17H24FN3O4 (exact mass: 353.18; MW: 353.39): (1H NMR, 13C NMR and mass spectra of compound 7d are shown
C, 57.78; H, 6.85; N, 11.89; O, 18.11; found: C, 57. 81; H, 6.66; N, in the ESI,† Fig. S10–S12 respectively).
12.1; O, 18.12% (1H NMR, 13C NMR and mass spectra of
compound 7b are shown in the ESI,† Fig. S4–S6 respectively).
N1-Butyl-2-(2-(4-chlorophenyl)acetamido)-N5-hydroxypentane-
diamide (7e). Yield 60%, cream color solid, H NMR (400 MHz,
1
2-(2-(4-Bromophenyl)acetamido)-N5-hydroxy-N1-pentylpenta- DMSO-d6) d 10.39 (s, 1H), 8.26 (d, J = 8.1 Hz, 1H), 7.88 (t, J =
nediamide (7c). Yield 48%, cream color solid, 1H NMR 5.6 Hz, 1H), 7.38–7.33 (m, 2H), 7.28 (d, J = 8.5 Hz, 2H), 4.22–4.12
(400 MHz, DMSO-d6) d 8.29 (d, J = 8.1 Hz, 1H), 7.90 (t, J = 5.5 Hz, (m, 1H), 3.48 (s, 1H), 3.03 (m, 1H), 1.97–1.60 (m, 2H), 1.40–1.17
1H), 7.47 (d, J = 8.4 Hz, 2H), 7.22 (d, J = 8.4 Hz, 2H), 4.17 (dd, J = (m, 7H), 0.86 (m, 3H). 13C NMR (101 MHz, DMSO-d6) d 171.25,
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
170.15, 168.91, 167.46, 135.91, 132.18, 132.06, 131.49, 131.38, (101 MHz, DMSO-d6) d 170.36, 168.44, 165.29, 135.30, 134.49,
131.25, 129.13, 128.54, 67.87, 52.77, 41.65, 38.58, 31.56, 30.26, 131.23, 130.51, 129.02, 127.11, 52.93, 38.49, 30.93, 28.95, 28.91,
29.48, 29.30, 28.95, 28.83, 23.71, 22.86, 19.92, 14.36, 14.11, 11.27. 27.96, 25.93, 22.02, 13.88. HPLC Rt: 2.810, LC/MS calculated for
HPLC Rt: 2.955 LC/MS calculated for expected C17H24ClN3O4 expected C18H25Cl2N3O4 [M]: 418.31; [M]+: 418, [M + 2]+: 420;
[M]: 369.84; Found: [M + H]+: 370, [M + Na]+: 392; anal. calcd for anal. calcd for C18H25Cl2N3O4 (exact mass: 417.12; MW: 418.31):
C17H24ClN3O4 (exact mass: 369.15; MW: 369.84): C, 55.21; H, C, 51.68; H, 6.02; N, 10.05; O, 15.30; found: C, 51.73; H, 5.99; N,
6.54; N, 11.36; O, 17.30; found: C, 56.09; H, 6.23; N, 11.42; O, 10.02; O, 15.49% (1H NMR, 13C NMR, and mass spectra of
16.83% (1H NMR, 13C NMR, and mass spectra of compound 7e compound 7i are shown in the ESI,† Fig. S25–S27 respectively).
are shown in the ESI,† Fig. S13–S15 respectively).
2-(2-(2,4-Dichlorophenyl)acetamido)-N5-hydroxy-N1-phenyl-
2-(2-(4-Chlorophenyl)acetamido)-N5-hydroxy-N1-pentylpenta- pentanediamide (7j). Yield 40%, off white powder, 1H NMR
1
nediamide (7f). Yield 80%, white crystals, H NMR (400 MHz, (400 MHz, DMSO-d6) d 8.75 (t, J = 9.0 Hz, 1H), 8.48 (t, J = 5.8 Hz,
DMSO-d6) d 8.31 (d, J = 8.1 Hz, 1H), 7.92 (t, J = 5.4 Hz, 1H), 7.34 1H), 7.68 (d, J = 1.2 Hz, 1H), 7.55–7.48 (m, 2H), 7.35–7.21 (m,
(d, J = 8.4 Hz, 2H), 7.28 (d, J = 8.4 Hz, 2H), 4.18 (dd, J = 13.7, 5H), 4.42 (m, 1H), 4.32 (d, J = 5.8 Hz, 2H), 2.15–1.95 (m, 3H),
7.9 Hz, 1H), 3.48 (s, 2H), 3.02 (m, 2H), 1.98–1.67 (m, 4H), 1.41– 1.91–1.80 (m, 1H), 13C NMR (101 MHz, DMSO-d6) d 170.74,
1.32 (m, 2H), 1.24 (m, 5H), 0.84 (t, J = 7.0 Hz, 3H). 13C NMR 170.72, 168.43, 165.45, 139.27, 139.23, 135.24, 134.53, 131.28,
(101 MHz, DMSO-d6) d 171.28, 170.19, 168.94, 135.90, 131.50, 130.56, 129.03, 128.23, 127.11, 127.04, 126.74, 53.06, 42.07,
131.38, 128.53, 52.81, 29.32, 29.10, 28.94, 22.24, 14.35. HPLC Rt: 28.94, 27.79. HPLC Rt: 2.712 LC/MS calculated for expected
2.362, LC/MS calculated for expected C18H26ClN3O4 [M]: 383.87;
C
19H19Cl2N3O4 [M]: 424.28; [M]+: 424, [M + 2]+: 426; anal. calcd
[M + H]+: 384, [M + Na]+: 406; anal. calcd for C18H26ClN3O4 for C19H19Cl2N3O4 (exact mass: 423.08; MW: 424.28): C, 53.79;
(exact mass: 383.16; MW: 383.87): C, 56.32; H, 6.83; N, 10.95; O, H, 4.51; N, 9.90; O, 15.08; found: C, 53.71; H, 4.80; N, 9.89; O,
16.67; found: C, 56.13; H, 6.7; N, 11.3; O, 16.27% (1H NMR, 13
C
14.98% (1H NMR, 13C NMR, and mass spectra of compound 7j
NMR, and mass spectra of compound 7f are shown in the ESI,† are shown in the ESI,† Fig. S28–S30 respectively).
Fig. S16–S18 respectively).
N5-Hydroxy-N1-pentyl-2-(2-phenylacetamido)pentanediamide
N1-Butyl-N5-hydroxy-2-(2-(4-nitrophenyl)acetamido)pentane- (7k). Yield 95.2%, white powder, 1H NMR (400 MHz, DMSO-d6) d
diamide (7g). Yield 57.9%, pale yellow powder, 1H NMR 8.39 (d, J = 8.2 Hz, 1H), 8.11–8.06 (m, 1H), 7.91 (m, 2H), 7.84–
(400 MHz, DMSO-d6) d 8.44 (d, J = 8.1 Hz, 1H), 8.17 (d, J = 8.7 Hz, 7.78 (m, 1H), 7.54–7.40 (m, 4H), 4.22 (m, 1H), 3.97 (d, J = 4.1 Hz,
2H), 7.95 (t, J = 5.5 Hz, 1H), 7.54 (d, J = 8.8 Hz, 2H), 4.19 (dd, J = 1H), 3.09–2.95 (m, 2H), 2.01–1.69 (m, 4H), 1.41–1.15 (m, 7H),
13.8, 8.2 Hz, 1H), 3.67 (s, 2H), 3.04 (m, 2H), 1.98–1.67 (m, 3H), 0.84 (t, J = 7.1 Hz, 3H), 13C NMR (101 MHz, DMSO-d6) d 170.79,
1.39–1.19 (m, 5H), 0.87–0.82 (m, 3H), 13C NMR (101 MHz, 169.96, 168.48, 133.28, 132.80, 131.95, 128.28, 127.70, 126.96,
DMSO-d6) d 170.69, 168.89, 168.41, 146.18, 144.55, 130.37, 125.85, 125.56, 125.43, 124.29, 52.31, 38.39, 28.87, 28.61, 28.52,
123.22, 52.39, 41.61, 38.09, 31.06, 28.81, 28.46, 19.42, 13.60. 28.45, 21.74, 13.85. HPLC Rt: 2.359, LC/MS calculated for
HPLC Rt: 3.388, LC/MS calculated for expected C17H24N4O6 [M]: expected C22H29N3O4 [M]: 399.48; [M + H]+: 400, [M + Na]+:
380.40; [M + H]+: 381, [M + Na]+: 403; anal. calcd for C17H24N4O4 422; anal. calcd for C22H29N3O4 (exact mass: 399.22; MW:
(exact mass: 383.16; MW: 383.87): C, 53.68; H, 6.36; N, 14.73; O, 399.48): C, 66.14; H, 7.32; N, 10.52; O, 16.02; found: C, 66.13;
25.24; found: C, 53.69; H, 6.4; N, 14.98; O, 24.99% (1H NMR, 13
C
H, 7.51; N, 10.39; O, 16.25% (1H NMR, 13C NMR, and mass
NMR, and mass spectra of compound 7g are shown in the ESI,† spectra of compound 7k are shown in the ESI,† Fig. S31–S33
Fig. S19–S21 respectively).
respectively).
N1-Butyl-2-(2-(2,4-dichlorophenyl)acetamido)-N5-hydroxypenta-
N1-Hexyl-N5-hydroxy-2-(2-phenylacetamido)pentanediamide (7l).
nediamide (7h). Yield 92%, cream powder, H NMR (400 MHz, Yield = 76.2%, white crystalline powder, 1H NMR (400 MHz,
DMSO-d6) d 8.29 (d, J = 8.1 Hz, 1H), 7.90 (t, J = 5.6 Hz, 1H), 7.57 (s, DMSO-d6) d 10.43 (s, 1H), 8.42 (d, J = 8.1 Hz, 1H), 8.12–8.05 (m,
1H), 7.38 (s, 2H), 4.21 (m, 1H), 3.65 (d, J = 3.0 Hz, 2H), 3.11–2.99 1H), 7.93–7.87 (m, 2H), 7.81 (dd, J = 6.3, 3.0 Hz, 1H), 7.53–7.41
(m, 2H), 2.01–1.69 (m, 3H), 1.41–1.21 (m, 5H), 0.86 (t, J = 7.3 Hz, (m, 4H), 4.24–4.17 (m, 1H), 3.98 (d, J = 2.8 Hz, 2H), 3.07–2.96
3H), 13C NMR (101 MHz, DMSO-d6) d 173.81, 170.69, 168.46, (m, 2H), 2.03–1.68 (m, 5H), 1.39–1.18 (m, 9H), 0.85 (t, J = 6.7 Hz,
134.53, 133.53, 133.13, 131.91, 128.34, 127.04, 52.34, 38.09, 31.08, 3H). 13C NMR (101 MHz, DMSO-d6) d 171.32, 170.49, 169.02,
30.10, 28.82, 28.45, 19.43, 13.62. HPLC Rt: 2.526, LC/MS calculated 133.77, 133.31, 132.45, 128.78, 128.22, 127.46, 126.35, 126.06,
for expected C17H23Cl2N3O4 [M]: 404.29; [M]+: 404, [M + 2]: 406; 125.93, 124.81, 52.87, 31.39, 29.40, 29.01, 26.30, 22.48, 21.54,
anal. calcd for C17H23 Cl2N3O4 (exact mass: 403.11; MW: 404.29): 14.37. LC-MS calculated for expected C23H31N3O4 [M]: 413.51
C, 50.50; H, 5.73; N, 10.39; O, 15.83; found: C, 50.68; H, 6.07; N, and found [M + H]+: 414; anal. calcd for C23H31N3O4 (exact
10.64; O, 15.13% (1H NMR, 13C NMR, and mass spectra of mass: 413.23; MW: 413.51): C, 66.81; H, 7.56; N, 10.16; O, 15.48;
1
compound 7h are shown in the ESI,† Fig. S22–S24 respectively).
2-(2-(2,4-Dichlorophenyl)acetamido)-N5-hydroxy-N1-pentyl-
pentanediamide (7i). Yield 88.6%, white powder, 1H NMR Fig. S34–S36 respectively).
(400 MHz, DMSO-d6) d 8.72 (s, 1H), 8.66 (d, J = 8.1 Hz, 1H),
found: C, 66.59; H, 7.67; N, 10.07; O, 15.67% (1H NMR, 13C
NMR, and mass spectra of compound 7l are shown in the ESI,†
HDAC inhibition assay and selectivity of compounds
7.91 (t, J = 5.6 Hz, 1H), 7.68 (s, 1H), 7.50 (d, J = 0.9 Hz, 2H), 4.34
(m, 1H), 3.14–3.01 (m, 2H), 2.12–1.75 (m, 4H), 1.39 (dd, J = 13.1, All target compounds were screened for their HDAC enzyme
6.4 Hz, 2H), 1.31–1.15 (m, 6H), 0.86 (t, J = 6.7 Hz, 3H), 13C NMR inhibition efficiency.16 First, we tested these compounds on
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
Fig. 4 HDAC inhibition efficiency of all target compounds. All novel
compounds were screened at 10 mM concentration on HeLa nuclear
extract, human recombinant HDAC3, and human recombinant HDAC8.
Fig. 5 Dose response curves of 7d, 7e, 7f, 7h, 7i, 7k, 7l, and SAHA in
HDAC8 inhibition assay [the data represents mean Æ SD (n = 3)].
HeLa nuclear extract, which is rich in HDAC1 and HDAC2
enzymes.36 It was observed that compounds 7c, 7d, 7e, 7f, 7h,
7i, 7k, and 7l were showing good inhibition (Fig. 4 and Table 4).
Again, they were tested on human recombinant HDAC8 at
the same concentration (10 mM). Almost all compounds were
found to show promising HDAC8 inhibition. Particularly,
compounds 7d, 7f, 7h, 7i, 7k, and 7l were showing more than
80% HDAC8 inhibition. Another Class I HDAC, HDAC3 was
also screened for inhibition by these compounds. These
compounds were found to show minimal inhibition of HDAC3.
These results demonstrate the selectivity of these compounds
towards the HDAC8 enzyme. More interestingly, these results
are in agreement with the prediction results from the MLR and
SVM models, as demonstrated in Table 4.
Regarding the HDAC enzyme inhibition, it was noticed that a
number of compounds (compounds 7c–7h, 7i, 7k, and 7l) showed
effective HDAC8 inhibition (450% at 10 mM). Among these,
compounds 7f, 7g, and 7k exhibited HDAC8 selectivity over HeLa
nuclear extract and HDAC3/NCoR1. On the other hand, the other
compounds are not very HDAC8 selective (Fig. 5).
Cell culture studies using anticancer cell lines
Antiproliferative assay using cancer cell lines (MTT assay).
Cytotoxicity studies were performed on three different cell
lines, i.e., B16F10 (murine melanoma cell line), non-small cell
lung cancer cell line (A549), and human acute T cell lympho-
blastic leukemia (Jurkat E6.1) by MTT assay method. These
synthesized compounds were evaluated at two different con-
centrations (100 and 10 mM) in triplicate (Fig. 6A–C).
From the assay results, it was observed that few compounds
displayed good cytotoxicity with a similar trend on these cell
lines. Promising compounds (7d, 7e, 7f, 7h, 7i) were further
screened to find out their IC50 values in all three cell lines.
Results are displayed in Fig. S37A–C in ESI† and Table 5.
Increased activity was observed with compounds having an
alkyl side chain with five carbon atoms. This length is the most
suitable as four and six carbon chains showed comparatively
lesser activity. Again, substitution on the phenyl ring showed
an impact on the activity. It was interesting to note that
Table 4 %HDAC inhibition of target compounds in HeLa nuclear extract, human recombinant HDAC3, and human recombinant HDAC8
% Inhibition at 10 mM
HeLa nuclear extract HDAC3/NCoR1 HDAC8 selectivity rangeb
Observed HDAC8/HDAC3 Average predicted HDAC8/
Compa R1
R2
HDAC3 selectivity rangec
7a
7b
7c
7d
7e
7f
7g
7h
7i
2-Br
2-F
4-Br
4-Br
4-Cl
4-Cl
4-NO2
n-C5H11 23.25
i-C4H9 19.14
n-C5H11 42.24
n-C6H13 67.16
16.71
4.65
9.02
11.96
7.73
12.67
11.36
13.63
20.33
22.11
10.10
11.08
38.84
40.17
60.24
83.14
69.66
78.84
50.61
82.05
89.17
39.44
86.38
78.94
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
SEL
n-C4H9
48.12
n-C5H11 41.46
n-C4H9
13.77
60.11
2,4-Dichloro n-C4H9
2,4-Dichloro n-C5H11 59.42
2,4-Dichloro C6H5 17.43
—
—
7j
7k
7l
n-C5H11 45.52
n-C6H13 51.69
a
b
c
Compound number. Observed selectivity (negative logarithmic ratio of HDAC8/HDAC3). Average predicted selectivity as per the QAAR models
shown for comparison; SEL, HDAC8 selective.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
Table 5 IC50 values of target compounds in different cancer cell lines
IC50 (mM) IC50 (mM) IC50 (mM)
inB16F10 in A549
in Jurkat
E6.1 cell line
Compa R1
R2
cell line
cell line
7a
7b
7c
7d
7e
7f
7g
7h
7i
2-Br
2-F
4-Br
4-Br
4-Cl
4-Cl
4-NO2
n-C5H11 450
i-C4H9 450
n-C5H11 450
n-C6H13 34.71
n-C4H9 38.16
n-C5H11 27.72
n-C4H9 450
450
450
450
31.47
38.03
20.33
450
33.39
22.54
450
450
450
450
450
450
28.66
34.42
18.57
450
30.08
17.24
450
450
450
2,4-Dichloro n-C4H9 32.21
2,4-Dichloro n-C5H11 28.13
2,4-Dichloro C6H5
—
—
7j
7k
7l
450
n-C5H11 450
n-C6H13 450
a
Compound number.
compound 7i may produce anticancer activity with the growth
arrest at the G2/M phase (Fig. 7).
Binding mode of interaction study
Since we earlier performed the binding mode of interaction studies
of carboxylic acid derivatives,8 we tried the same for hydroxamate
derivatives here. The molecular docking study37 revealed that all
these compounds were almost superimposed in the active site of
HDAC8 (PDB: 1VKG) (Fig. 8). The hydroxamate moiety of these
compounds closely resided with the Zn2+ ion (Fig. 8).
The molecular docking interaction of the two prototype
compounds (7i and 7f) showed that the carbonyl group of the
hydroxamate moiety had an interaction with the Zn2+ ion
(Fig. 9). Again, the carbonyl group of the hydroxamate moiety
accepts a hydrogen bond from Tyr306.
Fig. 6 MTT assay of the synthesized compounds on (A) B16F10, (B) A549,
and (C) Jurkat E6.1 cell lines. Cells were treated for 72 h, and cell viability
was measured by MTT reagent. Data represent mean Æ SD (n = 3).
Moreover, Tyr306 at the active site accepts hydrogen bonds
from the amide group attached with the n-pentyl moiety. The
4-chlorobenzyl group formed two p–p stacking interactions
with Phe152 and Phe208. Notably, two parallel-displayed p–p
interactions (F192 and F208) are also important signs for good
HDAC8 interaction.38 In the case of compound 7f, Tyr306 amino
acid residue accepts hydrogen bonds from the amide group
attached with the n-pentyl moiety. Tyr306 also donates hydrogen
bond to the carbonyl function of the hydroxamate moiety.
Interaction of these targeted compounds with Y306 (which is
not present in HDAC4)39 could be the reason for the observed
HDAC8 selectivity. Interestingly, the molecular docking study of
these compounds against other HDACs was not so encouraging.
compounds with bulky halogen substitution at the R1 position
(compounds 7d–7f, 7h, 7i) display cytotoxicity in the cancer
cell lines tested. Apart from that, the n-pentyl substitution at
the R2 position (compounds 7f, 7i) was found to be optimum
compared to the n-butyl (compounds 7e, 7h) and n-hexyl
substitution (compound 7d). Importantly, aryl substitution
(such as phenyl) at the R2 position yielded inactivity
(compound 7j). Again, naphthylacetyl derivatives (compounds
7k, 7l) were found to be ineffective.
Conclusion
Cell cycle analysis
We further investigated the effect of compounds 7f and 7i on cell This study presents the first QAAR model for investigating
cycle progression in the representative cancer B16F10 cell line (Fig. 7). structural requirements to attain HDAC8 selectivity over
From the result, it was found that compounds 7f and 7i HDAC3. Robust QSAR models were evolved from the linear
induced growth arrest at the G2/M phase of cell growth. Com- and non-linear techniques and were validated by means of
pound 7i induced significant growth arrest around 42.8% in the internal and external validation tests. The developed QAAR
G2/M phase compared to control, showing 9.32%, while G0/G1 models were utilized to design new molecules having HDAC8
phase decreased to 10.5% from 74.6%. A similar pattern of selectivity. Then, design compounds were synthesized and
growth arrest was found for 7f also. These results indicated that characterized. All target compounds were evaluated for their
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
Fig. 7 Cell cycle arrest induced in B16F10 cell line by (A) control, (B) 7f, (C) 7i.
These compounds were also tested for their antiproliferative
activity in three different cell lines (B16F10, HeLa, and
Jurkat E6.1) for murine melanoma, non-small cell lung carci-
noma (NSCLC), and acute lymphoblastic leukemia (ALL)
types of cancer. Promising compounds were also explored for
their IC50 calculation in these cell lines. Compounds 7f and 7i
were found to possess efficient anticancer potential in
MTT assay. Compound 7f and 7i induced significant cell
growth arrest in the G2/M phase, indicating good anticancer
potential in cell cycle analysis of representative cancer murine
melanoma B16F10 cell line. These compounds (7f and 7i)
showed a binding mode of interactions with the HDAC8
enzyme.
Notably, these compounds can be further modified to
improve their efficacy and selectivity. Likewise, one has to
consider the requisite structural features from the QAAR
models and then can predict the selectivity values using
reported QAAR models. Further, the developed models can be
used as efficient query tools for screening selective HDAC8
inhibitors. Together, this study offers an understanding of the
essential structural features or properties of the molecules for
proper binding in the active site of the HDAC8 enzyme and may
help to get better selective HDAC8 inhibitors as anticancer
agents in the future.
Fig. 8 The docked conformations of compounds (grey lines) superim-
posed with each other at the active site of HDAC8 (PDB: 1VKG).
HDAC enzyme inhibition efficiency with HeLa nuclear extract,
human recombinant HDAC3, and human recombinant HDAC8
enzymes. Interestingly, these compounds were found to be
effective HDAC inhibitors with preferential selectivity for
HDAC8. The five-carbon side chain was found to be the most
appropriate length for the linker moiety.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
Fig. 9 Molecular docking interactions of (A) compound 7i and (B) compound 7f with HDAC8 enzyme (PDB: 1VKG).
Experimental section
Design of molecules: QAAR studies
Two supervised learning approaches, namely multiple linear
regression (MLR) and support vector machine (SVM) were
used.29,30
To achieve robust quantitative activity–activity relationship
(QAAR) model, series of investigations, including dataset
preparation and descriptor calculations, were performed
followed by regression analysis with different multivariate
methods.
MLR based 2D-QAAR study was employed here to correlate
molecular descriptors with biological activity. The predictability
of the developed stepwise multiple linear regression (S-MLR)
model was evaluated using both internal and external validation
matrices.11,31–33
For the construction of the SVM model, we used John Platt’s
Sequential Minimal Optimization (SMO) algorithm with the
Waikato Environment for Knowledge Analysis (Weka) version
3.8.4.34–36 SVM is a learning approach based on the statistical
learning theory. Recently, this technique received huge success
in modeling biological and chemical properties. Here, the SVM
parameters (The complexity parameter C = 1.0, exponent
E = 1.0, epsilon parameter = 0.001, tolerance T = 0.001) were
used to construct the SVM model.11
Data set
A
dataset containing fifty-nine diverse molecules having
HDAC8 and HDAC3 inhibitory activities was collected from
the published work of the University of Illinois at Chicago17–26
to perform the QAAR studies. These compounds include
aliphatic and aromatic hydrocarbons, heteroaromatic, amide,
ester, keto, and sulphonyl compounds, as shown in Table S1 of
the ESI.†
Descriptor generation
Chemistry
The 2D structures of investigated compounds were drawn
by ChemDraw Ultra 8.0 software,27 and subsequently, the
molecular structures were subjected to geometry optimizations.
The energy minimizations were carried out by a molecular
orbital package (MOPAC) module with 100 iterations, and the
minimum RMS gradient was 0.10. Then, the 2D descriptors
were calculated from the freely available software ‘‘PaDEL-
descriptor’’.28 Initially, descriptors with constant values, which
did not cover the whole space, were deleted from the study to
decrease the redundancy.11
All starting materials and reagents were commercially available
and used without further purification. All reactions were
monitored by thin layer chromatography (TLC) using precoated
1
plates with silica gel F254 from Merck Millipore Co., USA. H
and 13C NMR spectra were recorded in DMSO-d6 using
Nuclear Magnetic Resonance (NMR) Spectrophotometer
High Performance-400 MHz (Bruker Biospin, Switzerland).
Mass spectroscopy was performed in LCMS-2020, Shimadzu
using ESI mode. The purity of these compounds was analyzed
by analytical LC/MS. Analysis was performed on an LCMS-8040
equipped with a photodiode array detector using a Shiseido
C18 4.6 Â 150 mm, 5m column at a flow rate of 1 mL/min with
the isocratic flow (20% A: 80% B; solvent A = water + 0.4% TFA,
solvent B = methanol).
Multivariate regression analysis
Multivariate regression was performed to correlate the
independent descriptors, which in our case are the ‘‘PaDEL-
HDAC8/HDAC3
50
descriptors’’, with the dependent variable (pIC
).
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
Synthesis of substituted pentanoic acid derivatives (6a–6l)
protocol. Briefly, 15 mL of diluted human recombinant HDAC8
enzyme (BML-SE145-0100) was incubated with 10 mL of
test inhibitor (10 mM), 25 ml of Fluor de Lyss HDAC8, and
deacetylase substrate (BML-KI178-0005). After 10 min
incubation, the reaction was terminated by adding 50 mL stop
solution containing Fluor de Lyss- Developer II (BML-KI176-
1250) and trichostatin A. The plate was incubated for 45 min at
37 1C, and fluorescence intensity was measured by using
Spectramax M4 (Molecular Devices, USA). All target compounds
were screened at 10 mM concentration in triplicate.
The substituted pentanoic acids (5a–5l) were dissolved in
anhydrous methanol and the flask was kept in an ice bath to
get 0 1C temperature (Scheme 1). Thionyl chloride (1.5–2 ml)
was added dropwise to the above mixture and stirred overnight.
The reaction mixture was concentrated and extracted with ethyl
acetate. The organic layer was washed individually with
aqueous saturated sodium bicarbonate and brine solutions
separately. The crude organic mixture was purified individually
by column chromatography using EtOAc : hexane (3 : 7) as the
mobile phase.
Cell culture study
Synthesis of substituted hydroxamic acid (7a–7l)
The murine melanoma cell line (B16F10) was cultured in
Dulbecco’s Modified Eagle Medium (DMEM) (Himedia
Laboratories Pvt. Ltd, Mumbai, India), non-small cell lung
cancer cells (A549) in Dulbecco’s Modified Eagle Medium/
Nutrient Mixture F-12 Ham (DMEM/F12, 1 : 1 mixture, Himedia
Laboratories Pvt. Ltd, Mumbai, India), and Jurkat E6.1 cell line
in Roswell Park Memorial Institute-1640 (RPMI-1640, Himedia
Laboratories Pvt. Ltd, Mumbai, India), supplemented with 10%
heat inactivated fetal bovine serum (Himedia Laboratories Pvt.
Ltd., Mumbai, India) and 1% of antibiotic solution (10 000 U
penicillin and 10 mg streptomycin per ml, Himedia Laboratories
Pvt. Ltd, Mumbai, India). Cells were cultured at 37 1C in a
humidified atmosphere with 5% CO2. Stock solutions of all
synthesized compounds were prepared in DMSO at a concen-
tration of 100 mM and stored.
The intermediate substituted esters (6a–6l) were dissolved in
methanol separately. 50% w/v aqueous hydroxylamine
(18.5 equivalents) was added individually to it and stirred for
30 min at room temperature. To the above mixture, sodium
hydroxide (1 M, 2 equivalents) solution was added and stirred
for 1.5 h separately. The reaction mixture was concentrated and
poured in ice cold water individually. It was neutralized by 1 M
hydrochloric acid separately. The product was extracted with
ethyl acetate, dried over anhydrous sodium sulphate, and
evaporated to get the final product (7a–7l).
HDAC inhibition assay
HeLa nuclear extract assay. The HDAC enzyme inhibition
assay was performed as mentioned in our earlier publications
by using HDAC colorimetric assay kit (BML-AK501, ENZO life
MTT assay
sciences) following vendor’s protocol. Briefly,
5 mL of
HeLa nuclear extract (BML-KI137-0500), 10 mL of assay buffer The antiproliferative activity of these synthesized compounds
(BML-KI143-0020), and 10 mL of sample solution were added was determined as mentioned in our earlier publications by
per well in a microtiter plate. 25 mL of Color de Lyss substrate using an MTT assay. 5 Â 104 cells were seeded in 96 well plates
(BML-KI138-0050) was added, and the plate was incubated in and incubated overnight. Cells were treated with synthesized
an incubator for 30 min at 37 1C, which was terminated by compounds at two concentrations (100 and 10 mM) in
the addition of a 50 mL mixture of developer plus stop solution. triplicates and incubated for 72 h. 50 mL of 5 mg mLÀ1 3-(4,5-
The plate was incubated for 15 min at 37 1C, and absorbance dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT;
was measured at 405 nm. All synthesized compounds were Himedia Laboratories Pvt. Ltd, Mumbai, India) was added
screened at 10 mM concentration in triplicate.
and incubated for 4 h. Then Formazan crystals were dissolved
HDAC3/NCOR1 assay. HDAC3 enzyme inhibition assay was using DMSO and evaluated spectrophotometrically at 570 nm
performed using HDAC3/NCOR1 fluorimetric drug discovery kit and 650 nm using Spectramax M4 (Molecular Devices, USA).
(BML-AK531, ENZO life sciences) following vendor’s protocol.
For the determination of IC50 values of promising
Here, 10 mL of sample solution and 15 mL diluted HDAC3/NCOR1 compounds, same procedure was followed as described above,
complex solution (BML-KI574-0030) were added per well in where five compounds (7d, 7e, 7f, 7h, and 7i) were tested at 10
a microtiter plate. 25 mL Fluor de Lyss substrate solution different concentrations as 400, 200, 100, 50, 25, 12.5, 6.25,
(BML-KI177-0005) was added. The plate was incubated for 3.125, 1.562, and 0.781 mM.
15 min at 37 1C. Then, 50 mL of a mixture of Fluor de Lyss
For Jurkat E6.1 cell line, 2 Â 104 cells in 100 mL were seeded
developer II (BML-KI176-1250) and trichostatin A (BML-GR309- in 96 well plates and incubated overnight as mentioned in our
9090) was added per well and incubated for 45 min at 37 1C. earlier publications. Treatment concentrations were the same
The fluorescence intensity was measured at an excitation wave- for both experiments as described above for other cell lines.
length of 360 nm and emission wavelength of 460 nm using After 72 h incubation with the test compounds, cells were
Spectramax M4 (Molecular Devices, USA). All synthesized com- treated with 22 mL of Prestobluet (Invitrogen, Life technologies,
pounds were screened at 10 mM concentration in triplicate.
United States) reagent. The plate was incubated for 2 h, and
HDAC8 assay using human recombinant HDAC8. HDAC8 fluorescence intensity was measured at the excitation wavelength
inhibition assay was performed using HDAC8 fluorimetric drug of 560 nm and emission wavelength of 590 nm using Spectramax
discovery kit (BML-AK518, Enzo life science) following vendor’s M4 (Molecular Devices, USA).
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
Cell cycle analysis
CSIR, New Delhi, India, for providing Research Associateship
(RA) [FILE NO.: 09/096(0966)/2019-EMR-I, Dated: 28-03-2019].
Sanjib Das is thankful to AICTE, New Delhi, India, for awarding
National Doctoral Fellowship (NDF). Balaram Ghosh is also
thankful to CSIR, New Delhi, India, for financial assistance (FILE
NO.: CSIR- 37(1722)/19/EMR-II). Tarun Jha is thankful for the
financial support from RUSA 2.0 of UGC, New Delhi, India to
Jadavpur University, Kolkata, India. Sk. Abdul Amin thankfully
acknowledges Dr Shovanlal Gayen of Jadavpur University,
Kolkata, India, for his critical discussion. Sandip Kumar Baidya,
Saptarshi Sanyal, Suvankar Banerjee of the Department of
Pharmaceutical Technology, Jadavpur University, Kolkata, India,
are gratefully acknowledged for multimodal inputs. The authors
sincerely acknowledge the Department of Pharmacy, BITS-Pilani,
Hyderabad, India, and the Department of Pharmaceutical
Technology, Jadavpur University, Kolkata, India, for providing
the research facilities.
Briefly, 5 Â 105 B16F10 cells were seeded in 6 well plates and
incubated overnight at 37 1C. The next day, cells were treated
with 7d and 7i, respectively, and incubated for 72 h. Cells were
trypsinized, centrifuged, washed twice with PBS, fixed with 70%
ice-cold ethanol, and stored at À20 1C for 24 h. The cells were
centrifuged, washed, and resuspended in 500 mL PBS at pH 7.4
containing RNAse (100 mg mLÀ1) at room temp for 20 min.
Propidium iodide (PI) (50 mg mLÀ1 in PBS) was added to stain
cellular DNA and kept in the dark for 30 min. Flow cytometry
analysis was performed, and the percentage of cells in each phase of
the cell cycle was estimated using the IDEAS Software (version 6.0).
Binding mode interaction study
In order to understand the probable enzyme–ligand binding
interactions, molecular docking studies of the synthesized
compounds were carried out by the Glide module of Maestro
37
¨
v12.1, Schrodinger Inc. USA. Three-dimensional structure of
HDAC8 (PDB code: 1VKG) was downloaded from RSCB Protein
Data Bank. It was further processed using ‘‘Protein Preparation
Wizard’’ in Maestro v12.1. Generating states and refinement
steps were performed. It automatically added hydrogen atoms
and some essential bonds at the missing sites of the proteins.
The further refinement steps included optimization of hydrogen
bonded groups and restrained minimization using the default
force field OPLS_3e. After the optimization process, receptor grid
generation was processed to locate the binding pocket. Then,
extra precision (XP) docking studies were achieved, and subse-
quently, the analyses of the binding were performed. The con-
formations were ranked on the basis of scores and selected as
per their interactions with the enzymes as well as docking scores.
References
1 M. M. S. Hamoud, S. Pulya, N. A. Osman, Y. Bobde,
A. E. A. Hassan, H. A. Abdel-Fattah, B. Ghosh and
A. M. Ghanim, Design, synthesis, and biological evaluation
of novel nicotinamide derivatives as potential histone
deacetylase-3 inhibitors, New J. Chem., 2020, 44, 9671,
DOI: 10.1039/d0nj01274b.
2 T. C. S. Ho, A. H. Y. Chan and A. Ganesan, Thirty years of
HDAC inhibitors: 2020 insight and hindsight, J. Med. Chem.,
2020, 63, 12460–12484, DOI: 10.1021/acs.jmedchem.0c00830.
3 S. A. Amin, N. Adhikari and T. Jha, Diverse classes of HDAC8
inhibitors: In search of molecular fingerprints that regulate
activity, Future Med. Chem., 2018, 10, 1589–1602, DOI:
10.4155/fmc-2018-0005.
Author contributions
4 A. K. Halder, S. Mallick, D. Shikha, A. Saha, K. D. Saha and
T. Jha, Design of dual MMP-2/HDAC8 inhibitors by phar-
macophore mapping, molecular docking, synthesis and
biological activity, RSC Adv., 2015, 5, 72373–72386, DOI:
10.1039/c5ra12606a.
5 H. Chu, Q.-X. He, J. Wang, Y. Hu, Y.-Q. Wang and Z.-H. Lin,
In silico design of novel benzohydroxamate-based com-
pounds as inhibitors of histone deacetylase 6 based on
3D-QSAR, molecular docking, and molecular dynamics
simulations, New J. Chem., 2020, 44, 21201–21210, DOI:
10.1039/D0NJ04704J.
Sk. Abdul Amin: conceptualization, investigation, methodology,
validation, data curation, writing – original draft, writing – review
& editing. Prakruti Trivedi: methodology, validation, data cura-
tion, writing – original draft. Nilanjan Adhikari: validation, data
curation, writing – original draft. Routholla Ganesh: methodol-
ogy, validation, data curation. Dhanya Vijayasarathi: methodol-
ogy, data curation. Sanjib Das: formal analysis, data curation.
Balaram Ghosh: project administration, funding acquisition,
writing – review & editing. Tarun Jha: conceptualization, project
administration, writing – review & editing, final supervision.
6 A. Chakrabarti, I. Oehme, O. Witt, G. Oliveira, W. Sippl,
C. Romier, R. J. Pierce and M. Jung, HDAC8: A multifaceted
target for therapeutic interventions, Trends Pharmacol. Sci.,
2015, 36, 481–492, DOI: 10.1016/j.tips.2015.04.013.
7 Y. Asfaha, C. Schrenk, L. A. Alves Avelar, A. Hamacher,
M. Pflieger, M. U. Kassack and T. Kurz, Recent advances
in class IIa histone deacetylases research, Bioorg. Med.
Chem., 2019, 27, 115087, DOI: 10.1016/j.bmc.2019.115087.
8 S. Dutta, A. K. Halder, N. Adhikari, S. A. Amin, S. Das,
A. Saha and T. Jha, Synthesis, anticancer activity, structure-
activity relationship and binding mode of interaction
Conflicts of interest
Authors do not have any conflict of interest.
Acknowledgements
Sk. Abdul Amin sincerely acknowledges the Council of Scientific
and Industrial Research (CSIR), New Delhi, India, for awarding
the Senior Research Fellowship [FILE NO: 09/096(0967)/2019-
EMR-I, Dated: 01-04-2019]. Nilanjan Adhikari is grateful to the
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021

View Article Online
NJC
Paper
studies of substituted pentanoic acids, Future Med. Chem., 19 J. A. Bergman, K. Woan, P. Perez-Villarroel, A. Villagra,
2019, 11, 1679–1702, DOI: 10.4155/fmc-2018-0361.
E. M. Sotomayor and A. P. Kozikowski, Selective histone
deacetylase 6 inhibitors bearing substituted urea linkers
inhibit melanoma cell growth, J. Med. Chem., 2012, 55,
9891–9899, DOI: 10.1021/jm301098e.
9 S. A. Amin, N. Adhikari, T. Jha and S. Gayen, First molecular
modeling report on novel arylpyrimidine kynurenine monoox-
ygenase inhibitors through multi-QSAR analysis against Hun-
tington’s disease: A proposal to chemists!, Bioorg. Med. Chem. 20 A. P. Kozikowski, S. Tapadar, D. N. Luchini, K. H. Kim and
Lett., 2016, 26, 5712–5718, DOI: 10.1016/j.bmcl.2016.10.058.
10 A. M. Srour, N. S. Ahmed, S. S. Abd El-Karim, M. M. Anwar
and S. M. El-Hallouty, Design, synthesis, biological evalua-
tion, QSAR analysis and molecular modelling of new
thiazol-benzimidazoles as EGFR inhibitors, Bioorg. Med.
Chem., 2020, 28, 115657, DOI: 10.1016/j.bmc.2020.115657.
11 E. Bourguet, K. Ozdarska, G. Moroy, J. Jeanblanc and
M. Naassila, Class I HDAC inhibitors: Potential new epigenetic
therapeutics for alcohol use disorder (AUD), J. Med. Chem.,
2018, 61, 1745–1766, DOI: 10.1021/acs.jmedchem.7b00115.
12 O. J. Ingham, R. M. Paranal, W. B. Smith, R. A. Escobar,
D. D. Billadeau, Use of the nitrile oxide cycloaddition (NOC)
reaction for molecular probe generation: A new class of
enzyme selective histone deacetylase inhibitors (HDACIs)
showing picomolar activity at HDAC6, J. Med. Chem., 2008,
51, 4370–4373, DOI: 10.1021/jm8002894.
21 B. He, S. Velaparthi, G. Pieffet, C. Pennington, A. Mahesh,
D. L. Holzle, M. Brunsteiner, R. van Breemen, S. Y. Blond
and P. A. Petukhov, Binding ensemble profiling with photo-
affinity labeling (BEProFL) approach: Mapping the binding
poses of HDAC8 inhibitors, J. Med. Chem., 2009, 52,
7003–7013, DOI: 10.1021/jm9005077.
H. Yueh, T. Snyder, J. A. Porco Jr, J. E. Bradner and 22 Y. Chen, M. Lopez-Sanchez, D. N. Savoy, D. D. Billadeau,
A. B. Beeler, Development of a potent and selective HDAC8
inhibitor, ACS Med. Chem. Lett., 2016, 7, 929–932, DOI:
10.1021/acsmedchemlett.6b00239.
G. S. Dow and A. P. Kozikowski, A series of potent and
selective, triazolylphenyl-based histone deacetylases inhibi-
tors with activity against pancreatic cancer cells and Plas-
modium falciparum, J. Med. Chem., 2008, 51, 3437–3448,
DOI: 10.1021/jm701606b.
13 N. Adhikari, A. K. Halder, S. Mallick, A. Saha, K. D. Saha and
T. Jha, Robust design of some selective matrix metallo-
proteinase-2 inhibitors over matrix metalloproteinase-9 23 R. He, Y. Chen, Y. Chen, A. V. Ougolkov, J. S. Zhang,
through in silico/fragment-based lead identification and
de novo lead modification: Syntheses and biological assays,
Bioorg. Med. Chem., 2016, 24, 4291–4309, DOI: 10.1016/
j.bmc.2016.07.023.
D. N. Savoy, D. D. Billadeau and A. P. Kozikowski, Synthesis
and biological evaluation of triazol-4-ylphenyl-bearing his-
tone deacetylase inhibitors as anticancer agents, J. Med.
Chem., 2010, 53, 1347–1356, DOI: 10.1021/jm901667k.
14 A. Mukherjee, N. Adhikari and T. Jha, A pentanoic acid derivative 24 T. Y. Taha, S. M. Aboukhatwa, R. C. Knopp, N. Ikegaki,
targeting matrix metalloproteinase-2 (MMP-2) induces apoptosis
in a chronic myeloid leukemia cell line, Eur. J. Med. Chem., 2017,
141, 37–50, DOI: 10.1016/j.ejmech.2017.09.052.
H. Abdelkarim, J. Neerasa, Y. Lu, R. Neelarapu,
T. W. Hanigan, G. R. J. Thatcher and P. A. Petukhov, Design,
synthesis, and biological evaluation of tetrahydro-
isoquinoline-based histone deacetylase 8 selective inhibi-
tors, ACS Med. Chem. Lett., 2017, 8, 824–829, DOI: 10.1021/
acsmedchemlett.7b00126.
15 P. Trivedi, N. Adhikari, S. A. Amin, T. Jha and B. Ghosh,
Design, synthesis and biological screening of 2-
aminobenzamides as selective HDAC3 inhibitors with pro-
mising anticancer effects, Eur. J. Pharm. Sci., 2018, 124, 25 T. Y. Taha, S. M. Aboukhatwa, R. C. Knopp, N. Ikegaki,
165–181, DOI: 10.1016/j.ejps.2018.08.030.
H. Abdelkarim, J. Neerasa, Y. Lu, R. Neelarapu,
T. W. Hanigan, G. R. J. Thatcher and P. A. Petukhov,
Correction to design, synthesis, and biological evaluation
of tetrahydroisoquinoline-based histone deacetylase 8 selec-
tive inhibitors, ACS Med. Chem. Lett., 2019, 10, 1358, DOI:
10.1021/acsmedchemlett.9b00336.
16 P. Trivedi, N. Adhikari, S. A. Amin, Y. Bobde, R. Ganesh,
T. Jha and B. Ghosh, Design, synthesis, biological evalua-
tion and molecular docking study of arylcarboxamido piper-
idine and piperazine-based hydroxamates as potential
HDAC8 inhibitors with promising anticancer activity, Eur,
J. Pharm. Sci., 2019, 138, 105046, DOI: 10.1016/ 26 R. Neelarapu, D. L. Holzle, S. Velaparthi, H. Bai,
j.ejps.2019.105046.
M. Brunsteiner, S. Y. Blond and P. A. Petukhov, Design,
synthesis, docking, and biological evaluation of novel
diazide-containing isoxazole- and pyrazole-based histone
deacetylase probes, J. Med. Chem., 2011, 54, 4350–4364,
DOI: 10.1021/jm2001025.
17 A. S. Vaidya, R. Neelarapu, A. Madriaga, H. Bai,
E. Mendonca, H. Abdelkarim, R. B. van Breemen,
S. Y. Blond and P. A. Petukhov, Novel histone deacetylase
8 ligands without a zinc chelating group: Exploring an
’upside-down’ binding pose, Bioorg. Med. Chem. Lett., 27 ChemDraw Ultra 8.0, Cambridge Soft Corporation, USA,
2012, 22, 6621–6627, DOI: 10.1016/j.bmcl.2012.08.104. 2015.
18 A. P. Kozikowski, Y. Chen, A. Gaysin, B. Chen, M. A. D’Annibale, 28 C. W. Yap, PaDEL-descriptor: An open source software to
C. M. Suto and B. C. Langley, Functional differences in
epigenetic modulators-superiority of mercaptoacetamide-
calculate molecular descriptors and fingerprints, J. Comput.
Chem., 2011, 32, 1466–1474, DOI: 10.1002/jcc.21707.
based histone deacetylase inhibitors relative to hydroxamates 29 C. Nantasenamat, C. Isarankura-Na-Ayudhya, T. Naenna
in cortical neuron neuroprotection studies, J. Med. Chem., 2007,
50, 3054–3061, DOI: 10.1021/jm070178x.
and V. Prachayasittikul, Prediction of bond dissociation
enthalpy of antioxidant phenols by support vector machine,
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021
New J. Chem.

View Article Online
Paper
NJC
J. Mol. Graphics Modell., 2008, 27, 188–196, DOI: 10.1016/ 35 T. C. Smith and E. Frank, Introducing machine learning
j.jmgm.2008.04.005.
concepts with WEKA, Methods Mol. Biol., 2016, 1418,
30 C. Nantasenamat, A. Worachartcheewan, S. Prachayasittikul,
353–378, DOI: 10.1007/978-1-4939-3578-9_17.
C. Isarankura-Na-Ayudhya and V. Prachayasittikul, QSAR 36 L. A. Serebryannyy, C. M. Cruz and P. de Lanerolle, A role for
modeling of aromatase inhibitory activity of 1-substituted
1,2,3-triazole analogs of letrozole, Eur. J. Med. Chem., 2013,
69, 99–114, DOI: 10.1016/j.ejmech.2013.08.015.
nuclear actin in HDAC 1 and 2 regulation, Sci. Rep., 2016,
6, 28460, DOI: 10.1038/srep28460.
¨
¨
37 Schrodinger Suite, 2018. Schrodinger. LLC, New York, USA.
¨
http://www.schrodinger.com/glide.
31 QSAR tools, DTC laboratory, India; Software available at
´
https://dtclab.webs.com/software-tools (accessed on 23rd 38 M. Morgen, R. R. Steimbach, M. Geraldy, L. Hellweg,
March 2021).
P. Sehr, J. Ridinger, O. Witt, I. Oehme, C. J. Herbst-
Gervasoni, J. D. Osko, N. J. Porter, D. W. Christianson,
N. Gunkel and A. K. Miller, Design and synthesis of
dihydroxamic acids as HDAC6/8/10 inhibitors, Chem-
MedChem, 2020, 15, 1163–1174, DOI: 10.1002/
cmdc.202000149.
32 P. Ambure and K. Roy, Exploring structural requirements
of leads for improving activity and selectivity against CDK5/
p25 in Alzheimer’s disease: An in silico approach, RSC Adv.,
2014, 4, 6702, DOI: 10.1039/c3ra46861e.
33 Organisation for Economic Co-operation and Development,
¨
Principles for the validation, for regulatory purposes, 39 N. Upadhyay, K. Tilekar, N. Jansch, M. Schweipert, J. D.
of (quantitative) structure–activity relationship models.
Available at https://www.oecd.org/chemicalsafety/risk-
assessment/37849783.pdf.
Hess, L. Henze Macias, P. Mrowka, R. J. Aguilera, J. Y.
Choe, F. J. Meyer-Almes and C. S. Ramaa, Discovery of
novel N-substituted thiazolidinediones (TZDs) as HDAC8 inhi-
bitors: in silico studies, synthesis, and biological evaluation,
Bioorg. Chem., 2020, 100, 103934, DOI: 10.1016/
j.bioorg.2020.103934.
34 I. H. Witten and E. Frank, Data mining: Practical machine
learning tools and techniques, Morgan Kaufmann, San Fran-
cisco, 2nd edn, 2005.
New J. Chem.
This journal is © The Royal Society of Chemistry and the Centre National de la Recherche Scientifique 2021