Synthesis and binary QSAR study of antitubercular quinolylhydrazides

Article abstract of DOI:10.1016/j.bmcl.2013.06.076

In continuation with our previous work in anti-TB research area, in the present study we have demonstrated the structural diversity of quinolylhydrazides as potent anti-tuberculars. The compound library was synthesized by molecular hybridization approach and tested in vitro against Mycobacterium tuberculosis H₃₇Rv strains. Among the designed conjugates, the most promising molecules were found to exhibit 100% Growth Inhibition (GI) at MIC <6.25 μg/mL. Moreover, several analogs in the designed series were also turned out as excellent anti-tuberculars. To probe the structural characteristics influencing on the SAR, the classification model was generated using a binary QSAR approach termed recursive partitioning (RP) analysis. The significant features outlined by the RP model act as a guide in order to design the 'lead' compound.

Full text of DOI:10.1016/j.bmcl.2013.06.076

Accepted Manuscript
Synthesis And Binary QSAR Study of Antitubercular Quinolylhydrazides
Atul Manvar, Vijay Khedkar, Jignesh Patel, Vipul Vora, Narsinh Dodia, Gautam
Patel, Evans Coutinho, Anamik Shah
PII:
S0960-894X(13)00809-3
DOI:
http://dx.doi.org/10.1016/j.bmcl.2013.06.076
BMCL 20640
Reference:
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date:
Revised Date:
Accepted Date:
14 March 2013
8 June 2013
27 June 2013
Please cite this article as: Manvar, A., Khedkar, V., Patel, J., Vora, V., Dodia, N., Patel, G., Coutinho, E., Shah, A.,
Synthesis And Binary QSAR Study of Antitubercular Quinolylhydrazides, Bioorganic & Medicinal Chemistry
Letters (2013), doi: http://dx.doi.org/10.1016/j.bmcl.2013.06.076
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Synthesis And Binary QSAR Study of Antitubercular
Quinolylhydrazides
Atul Manvara,b, Vijay Khedkard, Jignesh Patela, Vipul Voraa, Narsinh Dodiaa, Gautam Patelc,
Evans Coutinhod* and Anamik Shaha*
aNational Facility for Drug Discovery, Department of Chemistry, Saurashtra University, Rajkot-360005, Gujarat,
India.
bPresent address: Centre for Synthesis & Chemical Biology, UCD School of Chemistry & Chemical Biology,
University College Dublin, Belfield, Dublin 4, Ireland.
cHDH Pharma Inc., Morrisville, NC-27560.
dDepartment of Pharmaceutical Chemistry, Bombay College of Pharmacy, Kalina, Santacruz (E), Mumbai 400098,
India.
*Corresponding authors:
Anamik Shah
Tel/fax: +91 281 2581013/2589609; Email: anamik_shah@hotmail.com
https://sites.google.com/site/anamikkantilalshah/home; http://chemdive.com/
Evans Coutinho
Tel: +91 2226670905, Email: evans@bcpindia.org
1

Abstract
In continuation with our previous work in anti-TB research area, in the present study we have demonstrated
the structural diversity of quinolylhydrazides as potent antituberculars. The compound library was synthesized by
molecular hybridization approach and tested in vitro against M. tuberculosis H37Rv strains. Among the designed
conjugates, the most promising molecules were found to exhibit 100% Growth Inhibition (GI) at MIC <6.25 µg/mL.
Moreover, several analogs in the designed series were also turned out as excellent antituberculars. To probe the
structural characteristics influencing on the SAR, the classification model was generated using a binary QSAR
approach termed recursive partitioning (RP) analysis. The significant features outlined by the RP model act as a
guide in order to design the „lead‟ compound.
Keywords: Quinolones, Hydrazides, Mtb, H37Rv, MABA, Binary QSAR
Mycobacterium tuberculosis (Mtb), the pathogen remains one of the most fatal infectious
disease followed by the AIDS.1-3 According to WHO statistics one third of the world‟s
populations have been exposed to TB bacterium and new infections occurs at a rate of one per
second.4-7 Control and prevention of TB is a major task nowadays. Despite the need for better TB
therapies no new clinical candidates have been developed in the last few decades.8,9 Currently,
TB is treated with a schedule of four drugs combination (e.g. isoniazide, rifampin, pyrazinamide,
and ethambutol) and the treatment lasting up to 6-9 months.10,11 There are several demerits of the
existing treatment, such as exceedingly lengthy therapy, host toxicity, ineffectiveness against
resistance strains etc.12 Moreover, the drugs utilized for curing TB shows potential side effects
such as thrombocytopenia, neuropathy, rashes fever, and drug induced hepatitis.13 Furthermore,
the evolution of its new virulent forms, multidrug resistant (MDR-TB) and extensively drug
resistant (XDR-TB) have become a major threat of mankind.14 The XDR-TB is virtually
untreatable using current therapeutics and without strengthening of the current TB controls
measures.15 The significant challenges for TB control are increasing number of
immunocompromized individuals with HIV infections, who are highly susceptible to the
disease.16 Consequently, there is a pressing need to develop a novel, potent, and fast-acting
antituberculosis drugs having minimal toxicity profile that would administered in conjunction
with antireteroviral drugs.17,18 Currently, the global TB development pipeline has nine drugs.19
Furthermore, various quinoline based motifs are also at preclinical and clinical stages for TB
drug development. The fluoroquinolones, such as gatifloxacin and moxifloxacin targets DNA
topoisomerase IV and DNA gyrase can be utilized as anti-TB agents, however, they often suffer
from the resistance. The TMC207 is a highly potent anti-TB agent which is currently in phase II
clinical trials (Figure 1).20-22
2

Figure 1: Quinoline based scaffolds at various stages for TB drug development (1).
In this context our group has investigated a wide range of heterocycles23-27 as promising
antituberculars. Among them the 3D-QSAR study and other in silico analysis based rational
design of tetrahydropyrimidine24 was identified as a „lead‟ molecule. Inspired by these findings,
in the current paper, we have envisioned for the synthesis of quinolylhydrazides as potent anti-
TB agents by fragment based approach.
Scheme 1: Synthesis of 3-(2-phenylhydrazono)quinoline-2,4-(1H,3H)-dione derivatives.
At the outset, the synthetic strategy for the preparation of molecularly diverse hydrazides
of quinolones is depicted in Scheme 1. The target molecules were prepared by straightforward
Meerwein arylations.28 The 4-hydroxyquinolones bearing electron-rich, electron-deficient as well
as sterically hindered substituents were efficiently coupled with the corresponding diazonium
3

salts at lower temperature, afforded compounds 4-82 in moderate to good yields with a very high
functional group tolerance (see supporting informations). The antimycobacterial activities of
synthesized derivatives were assessed by employing the microplate alamar blue assay (MABA)
against Mtb H37Rv strains29 utilizing rifampicin as a reference drug.
Table 1: In vitro activity of compounds 4-82 against M. tuberculosis H37Rv strains.
aGI
(%)
bMIC
(μg/mL)
aGI
(%)
bMIC
(μg/mL)
Substitutions
Ring A
4-Br
2-Ph
H
Substitutions
Ring A
4-F
2,3-di Me
3-Cl
4-F
2-Me, 4-OMe 2-NO2
3-Me
3-Cl
2-Me, 4-OMe 2,5-di Cl
2-Me, 4-OMe 2-NO2
3-Cl
Compds
cCLogP Compds
cCLogP
R
H
H
H
H
H
H
H
H
H
H
Me
H
H
H
H
Me
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
Ring B
3-NO2
3-NO2
2-Cl
R
H
H
H
H
H
H
H
H
Me
H
Me
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
Me
Me
H
H
H
H
H
Me
H
H
Ring B
3-Cl
2-NO2
3-Cl
4
5
6
7
8
9
100
100
100
100
99
99
98
98
98
97
97
96
96
95
95
94
92
91
89
89
85
85
82
82
80
80
80
79
76
76
71
70
65
64
62
61
60
56
55
54
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
<6.25
2.643
3.668
2.750
1.923
1.780
2.493
2.728
3.463
2.782
2.037
3.412
1.699
2.493
3.035
4.056
3.412
2.279
2.493
3.086
1.219
3.343
2.636
3.249
4.056
2.279
3.468
3.606
3.086
2.750
3.035
2.893
2.636
1.523
2.493
3.842
2.493
1.523
2.992
2.669
3.842
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
50
48
47
46
45
42
41
40
40
39
38
37
36
35
34
32
32
31
31
29
27
24
24
22
21
21
20
20
19
14
13
12
11
10
7
<6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
>6.25
2.893
2.728
3.463
1.923
2.198
2.536
2.750
3.881
3.074
3.249
3.626
1.699
3.382
3.668
3.249
4.176
1.699
3.035
3.249
2.189
3.086
2.189
2.493
2.778
4.056
3.206
2.656
2.352
2.279
3.262
3.748
3.168
3.962
4.125
3.463
2.189
2.669
2.493
1.699
4-F
H
4-Cl
2-Me, 3-Cl
4-Cl
3-NO2
2-NO2
2-NO2
3-NO2
4-Cl
2-NO2
H
H
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
4-COMe
H
2, 3-di Cl
H
2-Me
H
2-NO2
2,5-di Cl
2-NO2
2-Cl
2,5-di Cl
2-NO2
2-Me
3-Me
4-Cl
2,3-di Cl
2-Cl
2-Cl
2-NO2
2-Cl
3-NO2
H
2-Me
2-OMe
3-NO2
3-Me
3, 4-di Cl
2-NO2
4-Me
4-NO2
4-NO2
4-COMe
H
3-Cl, 4-F
3-Cl
3-Cl
2-NO2
3-F
4-F
H
3-NO2
4-Cl
2-NO2
4-Cl
H
4-Cl
4-OMe
4-OMe
2-Ph
4-Me
3-Cl
2-OMe
3-Me
2-Cl
4-COMe
3-NO2
4-COMe
3-NO2
2,6-di Me
3,4-di Cl
2,3-di Cl
4-NO2
4-COMe
3-Me
4-OMe
2,6-di Me
2-Me, 4-OMe 4-Cl
4-Me
4,5-di Cl
2-Cl
3-NO2
4-Cl
2, 3-di Cl
2-NO2
2, 3-di Cl
2-NO2
2-NO2
2, 3-di Cl
2-NO2
2, 3-di Cl
2, 5-di Cl
2-NO2
3-Cl
3, 4-di Cl
H
H
3-NO2
2,3-di Cl
4-Cl
3-Me
4-F
3-Me
4-Cl
4-Cl, 5-F
4-NO2
2-NO2
3-Me
4-Cl
3-NO2
2-Me, 3-Cl
4-OMe
4-Me
3-NO2
2-NO2
4-Cl
2, 3-di Cl
2-NO2
2-NO2
2-NO2
4-Cl
2,5-di Cl
H
2-Cl
3-Cl
3-Cl
3-Cl
4-OMe
4-COMe
4-OMe
4-NO2
H
6
4
1
0
H
H
H
2, 3-di Cl
aGrowth Inhibition (GI) of virulent strains of M. tuberculosis; MIC of Rifampicin: 0.015–0.125 μg/mL against M.
tuberculosis H37Rv (97% inhibition); cCLogP is calculated on ChemDraw Ultra 12.0.
b
4

The structure activity relationship (SAR) of the molecular diversity is summarized in Table 1.
The GI values were obtained in the range of 0-100% at MIC <6.25 and >6.25 µg/mL. The
presence of versatile substituents on quinolylhydrazide skeleton was significantly affected on
antitubercular activity. The unsubstituted analog 13 was found to be considerably potent (97%
GI at MIC <6.25µM). Next, the influence of hydrophilic or lipophilic variants on the parent
structure was probed. The analogs in which ring B constituted with NO2 group, compounds 4, 6-
10, 15, 17, and 20 displayed superior activity. The compound 28 has shown 80% GI while, 36
and 40 exhibited almost 2-fold reductions in anti-TB activity. In the same context, the NO2
containing quinolylhydrazides, for instance 41, 45, 47, 48, 55, 57, 60, 67, 69, and 72 were found
to be weakly actives. On the contrary, the presence of NO2 group in ring A, in lieu of ring B,
among the compounds 16, 21, 22, 37, 64, 66, and 81, only few analogs, such as 16, 21, 22
exhibited good activity. This trend has clearly indicated that not only the presence of NO2 but its
site-selectivity in the particular ring is also essential to generate a „lead‟ molecule. Among N1
methylated analogs, compounds 14 and 19 were proven to be actives. While, in comparison
between 48 and 52 suggested that the N-methylation have not essential to design the potent
antitubercular. We next scrutinized di-halogenated surrogates 11, 18, 27, 46, 59, 68, and 78 have
shown GI ranging from 7-98%, nevertheless CLogP were observed <5. The analogs 11 and 18
were found to be notably promising, while up to 4-fold reduction in anti-TB activity was
observed within rest of the halogenated analogs. The incorporation of fluorine group may alter
many of the physical properties of organic molecules, such as lipophilicity, metabolic stability,
and conformational characteristics. For these reasons, we have rationally incorporated F group in
conjunction with Cl either in ring A or B in compounds 29, 30, 34, and 44, nevertheless, no
profound biological activities were noticed. Consequently, we may conclude that the blending of
F and Cl was proven to be crucial in a search of prominent antitubercular of this class. In
comparison between two sterically hindered analogs, compound 5 demonstrated 100% GI while,
dramatic loss in activity was noticed in other regio-isomer 57. When both rings comprising with
Me group, compound 33 has shown satisfactory reduction in activity while 61 was found to be
weakly potent. The molecules 38, 42, 43, 53, 62 exhibited moderate to poor activity. Notably
promising compound 12 bearing COMe and Cl groups in rings A and B respectively. The least
potent analog in the series, compound 82 constituted with OMe group in ring B, nevertheless
CLogP is 1.699. In comparisons, the molecules 23, 63, 65, and 79 in which 4-COMe in ring A
5

and NO2 or Cl in ring B, a mere analog 23 was found active, whilst the biological profile of rest
of the analogs were reduced up to 4-fold. The combination of OMe or Me and Cl, compounds
56, 58, 73-76 were found weakly potent. Finally, the effect of compound polarity was estimated
by calculating the CLogP. The thumb rule for CLogP to a drug-like molecule must be less than
„5‟ to by-pass a cell barrier. The CLogP seems to correlate some extent with lipophilicity and
was found in the range of 1.219-4.176. Despite the lower potency, none of the analog has shown
CLogP less than 5. Thus, from the SAR results we may confirm that the influences of electron
density of the substituents, lipophilicity, as well as stereo-electronic properties are indispensable
to design a „lead‟ structure.
Next, to classify this dataset, a binary QSAR utilising the recursive partitioning (RP)
analysis was performed. The RP is a simple statistical data analysis technique that seeks to
decipher the elusive relationships in dataset involving thresholds, interactions, and non-
linearities. The RP has an advantage of classifying biological activity data by considering an
appropriate descriptors recursively. The non-linearity in the data impede the analysis that is
based on linearity such as multiple linear regression (MLR), principal component analysis (PCA)
or partial least squares (PLS) regression. The RP analysis is inherently faster than other grouping
techniques for instance, clustering and the 2D nature makes facile for the data which is difficult
to sieve in to usable divisions of classification. Furthermore, it has to be pointed out that the RP
analysis is meant to identify the common characteristics of the binding modes in the data set and
to recognize other compounds those fit the acquired rules.30 The tree display of the SAR results
help to identify the decisive factors that may categorize the dataset in to molecular classes with
higher and lower potency. The algorithm accounts for all possible binary splits for each
descriptor and to identify the optimal by splitting criterion–a Boolean function BN(M) of the
molecular descriptors, returning „true‟ or „false‟. The resulting subsets recursively split further
with every new subset representing a node at which the splitting criterion is associated. The
branches of the tree signify the truth values taken by the splitting criteria. The true response to
any given splitting criteria follows a branch to the downside while the false response follows a
branch to the upside in the partition tree. There can be several nodes consisting only „active‟
molecules each representing different structural classes of known active. The goal is to produce
the RP model with decision tree in which the final terminal nodes contain either only „active‟ or
6

only „inactive‟ molecules or in other words to produce terminal subsets of maximal homogeneity
with respect to the measured activities of their members.
We probed a set of 79 molecules (compounds 4-82) for the RP analysis. The
classification model was generated using the CSAR recursive partitioning technique incorporated
in Cerius2 (version 4.8, Accelrys Inc)31 software. The numerical descriptors that encode
topological, geometric/structural, electronic, and thermodynamic properties were considered for
deriving the RP model32 and decision tree was generated from the results of the RP. The
antitubercular activities of the molecules under study were described in terms of their growth
inhibition (GI) values ranging from 0-100%. The entire dataset was split in to two classes as
inactive (0) and active (1) with the class 1 containing 38 less active compounds having GI values
<50% while the class 2 consisting 41 active compounds having GI values ≥50%. The correlation
matrices were built for the descriptors and the descriptors with zero variance as well as columns
containing 95% of zero values were eliminated. Moreover, the descriptors having cross-
correlation coefficient >0.5 were eliminated as they have represented nearly same information.
The resulting uncorrelated descriptors were used as independent variables (X) while the
biological activity served as the dependent variable (Y) for the RP analysis. The activity classes
were weighted equally (weighing by classes) wherein an active node is a node in which the
fraction of actives exceeds the fraction of active molecules over the entire dataset (in case of
weighing by observables a node with more than half of the members being active will be
considered „active‟). The inactives categorized in an „active‟ node while the actives classified in
„inactives‟ nodes are accounted as „misclassified‟. The splits were scored using Twoing rule
scoring function to minimize the misclassification costs; the pruning factor was varied from 0 to
3; the values for maximum tree depth were varied between 5 and 10 while default values were
used for maximum number of generic splits (30) and the number of knots per variable (20). The
classification model was validated using cross validation, with the number of cross-validation
groups set to 5.
The classification model was developed using 2D as well as 3D-physicochemical
descriptors by varying the parameters discussed above seeking to improve the following
parameters: „Class % Observed Correct‟, a measure of intra-class prediction accounting for the
compounds predicted correctly to be in respective classes as a percentage of a total number of
compounds observed to be in each class; ‘Overall % Predicted Correct’, so-called overall
7

prediction is the percent of ratio of a total number of compounds correctly classified to the
number of compounds predicted in that class providing information on the accuracy of an overall
prediction; the ‘enrichment factor’ for a specific class is the ratio of percentage of compounds
correctly predicted belong to that class over the original percentage of compounds belonging to
that class. The statistical results of the RP analysis are summarized in Table 2.33
Table 2. Statistical results of recursive partitioning model.
Number of
Compounds
38
Class %
Obs Correct Pred Correct
Overall %
Class Value
%
Enrichment
1
2
0
1
48.10
51.90
100
100
100
100
2.08
1.93
41
The results of the RP analysis were displayed as decision tree derived from the RP
process splitting the whole dataset in to smaller subsets (nodes). The RP progresses sequentially
examining all the descriptor variables to identify the best criterion for splitting the dataset in to
„active‟ or „inactive‟ class. The 18-leaf RP decision tree was obtained for the quinolylhydrazides
as depicted in Figure 2. The decision tree looks like the horizontal dendrogram formed with 18
terminal and 17 non-terminal nodes in which Class 1 (0) is plotted using Red while Class 2 (1)
using Green. The terminal nodes 1, 3, 5, 7, 10, 12, 15, 17, and 18 represents class 2 (i.e. active)
while the terminal nodes 2, 4, 6 8, 9, 11, 13, 14, and 16 represents class 1 (i.e. inactive). The
analysis of decision tree shows that in class 1, i.e. the class of „inactives‟, 38 out of the 38
compounds were correctly classified as belonging to class 1. For class 2 which is the class of
„actives‟, all 41 compounds were correctly classified as actives represented 100% intra-class
predictivity. The overall prediction was represented by „Overall % Predict Correct‟ was found to
be 100% signified the accuracy of prediction when entire data set was being predicted with the
RP model. The enrichment factor (2.08 for class 1 and 1.93 for class 2) also indicates that the RP
model is statistically significant and can be used to classify new compound library. To evaluate
the predictability and to avoid over fitting of data, the model was subjected to 5-fold cross-
validation. This procedure leaves out 5% of the dataset and builds the model using remaining
molecules which are then utilized to predict the activity of eliminated subset. The statistics of
cross-validation are depicted in Table 3.
8

Figure 2. An 18-leaf decision tree derived from the Recursive Partitioning (RP) for prediction of antitubercular
activity classes. Those marked 1:1, 2:0, 3:1, 4:0, 5:1, 6:0, 7:1, 8:0, 9:0: 10:1, 11:0, 12:1, 13:0, 14:0, 15:1, 16:0, 17:1
and 18:1 correspond to terminal nodes 1-18 and each terminal node corresponds to the value of 1 (active) or 0
(inactive).
Table 3. Statistical results of recursive partitioning model for 5-fold cross-validation.
Number of
Compound
38
Class %
Obs Correct Pred Correct
Overall %
Class Value
%
Enrichment
1
2
0
1
48.10
51.90
52.63
60.98
55.56
58.14
1.16
1.12
41
9

The acceptable percentage of classification, 56% and 58% was accomplished for class 1 and
class 2 with enrichment factor of 1.16 and 1.12, respectively indicated that the classification
model would be able to classify and predict the activity (class) of new derivatives. The
statistically significant cross-validation experiment signals for the given data set, a series of n
partition trees can be derived based on subsets of size (1 - l/n), which correctly represents active
apart from inactive in remaining l/n dataset. The key elements identified in the validation set
were consistent with those from the entire set.
The first primary split was observed on the density (a spatial descriptor reflecting the type
of atoms and how tightly they are packed in a molecule and also signify transport and melt
behavior) and its cut-off value was 1.24. The dataset was split in to two branches-those
molecules with density value less than 1.24 (compound 33) followed a branch to downside while
those (compound 46) with density value above cut-off followed a branch to upside. The
molecules with density greater than 1.24 were divided on the basis of area at the cut-off value of
342.76 with molecule (compound 6) having an area greater than the cut-off values formed the
branch to upside while those (compound 40) with area less than 342.76 followed a branch
downside. Six molecules with area greater than 342.76 were further split in to two branches on
the basis of atomic polarizability (Apol). It is an electronic descriptor and is a measure of the
relative tendency of charge distribution caused by the presence of nearby ion or dipole. It is also
related to hydrophobicity and thus to the biological activity. The molecule 12 with Apol greater
than 15744.63 formed the terminal node 1 and was correctly predicted as active, while the
molecules 51, 56, 64, 73 and 76 formed the terminal node 2 was correctly classified as inactive
by the RP model. Forty molecules with area less than 342.76 were split based on Dipole moment
(Dipole-mag) with cut-off value of 9.96. The dipole moment is an electronic descriptor
signifying the strength and orientation behavior of the molecule in an electrostatic field which
can be related to the long-range ligand-receptor recognition and subsequent binding. Ten
molecules 16, 18, 21, 30, 50, 59, 66, 68, 77, and 81 with Dipole-mag greater than 9.96 followed
a branch upside while those below cut-off formed a branch downside. These molecules were
further split using the Dipole-mag with molecules having values greater than 11.72 formed the
terminal node 3 while, the remaining seven molecules followed a branch downside. The terminal
node 3 contains the three molecules 16, 18, and 21 which were correctly classified as actives.
The remaining seven molecules with Dipole-mag less than 11.72 but greater than 9.96 were
10

further split to form the terminal nodes 4 and 5 on the basis of BIC (Bonding Information
Content). The molecules 50, 59, 66, 68, 77, and 81 with BIC values greater than 0.69 formed the
terminal node 4 and were predicted as inactive while molecule 30 having BIC less than 0.69 was
correctly classified as active and formed the terminal node 5. Thirty molecules with Dipole-mag
less than 9.96 were formed the branch downside were further split on the basis of IC
(Information Content) values. The molecule 69 with IC value greater than 3.88 formed the
terminal node 6 was correctly predicted as inactive by the RP model while, the remaining 29
molecules with IC values less than 3.88 were further split on the basis of Radius of Gyration with
cut-off of 4.42. Nineteen molecules with Radius of gyration (a spatial descriptor related to the
size of a molecule) greater than 4.42 formed the terminal node 7 while, the remaining 10
molecules with the descriptor value less than cut-off followed a branch downside. These were
again split on the basis of radius of gyration with the molecules having radius of gyration greater
than 4.41 but less than 4.42 formed the terminal node 8 and was found to contain two molecules
46 and 47 correctly predicted to be inactive. Eight molecules with Radius of Gyration less than
4.41 were now split using the descriptor variable density as the splitting criterion with molecule
78 having the density greater than 1.30 formed the terminal node 9 was correctly predicted as
inactive. The remaining seven molecules 6, 24, 28, 32, 34, 37, and 44 with density less than 1.30
formed the terminal node 10 and were correctly classified as active.
Switching to the first primary split thirty three molecules with density value less than
1.24 followed a branch downside. These molecules were further split in to two branches based
on dipole-mag property. The molecules 5, 15, and 20 with Dipole-mag values less than 3.61
formed the terminal node 18 and were correctly predicted as belonging to active class. The
remaining 30 molecules followed a branch upside and were split on the basis of density with cut-
off value as 1.12. The compounds 14 and 33 with density less than 1.12 formed the terminal
node 17 while other twenty eight molecules followed a branch upside in the partition tree. These
were further split on the basis of BIC with 5 molecules having the property value less than 0.7
followed a branch downside while those having BIC value greater than 0.7 followed a branch
upside. A branch having molecules with BIC less than 0.7 was split in to two sub-nodes on the
basis of Dipole-mag. The molecules 13, 19, and 23 having the Dipole-mag values greater than
5.69 formed the terminal node 15 while, the molecules 57 and 67 with Dipole-mag property less
than cut-off value formed the terminal node 16 and were correctly predicted by the RP model. In
11

viewing 23 molecules, out of 28 which showed the BIC value greater than 0.70 were split based
on their density variable. Fifteen molecules with density values less than 1.21 formed the
terminal node 14 and were correctly classified as inactive. The remaining eight molecules with
density greater than 1.21 followed a branch upside and were split on the basis of molecular
volume (Vm). It is a spatial descriptor representing the molecular volume inside the contact
surface. It is calculated as a function of conformation and is associated with binding and
transport of a molecule. Three molecules 17, 42, and 63 with molecular volume less than 267.56
followed a branch downside and those with descriptor value greater than 267.56 formed the
terminal node 11 containing the correctly predicted inactive compounds 55, 60, 65, 79, and 80.
The molecules 17, 42, and 63 were further split based on strain energy with correctly predicted
active molecules 17 and 42 forming the terminal node 12 while molecule 63 having strain energy
less than 8.15 formed the terminal node 13. Thus, the RP approach is sequentially dividing the
dataset on the basis of their physico-chemical descriptor, which help to identify the variables
need to be modified in order to arrive at the active molecule.
In conclusion, we have demonstrated the synthesis as well as antimycobacterial screening
of a new family of hydrazide. A significant numbers of compounds were found highly potent in
the preliminary screening assay. Furthermore, the recursive partitioning model derived for the
quinolylhydrazides encodes the useful information pertaining to chemical environment around
the molecules. The computational study predictions were also found to be in harmony with
antimycobacterial activity data. Despite a high degree of structural similarity in the dataset, the
classification model has been able to break down the large dataset in to a series of subsets those
are enriched in either active or inactive molecules. This model will be evolved continuously by
enlarging the dataset to generate a better hypothesis for designing novel candidates targeting
tuberculosis. Further lead optimization and detailed biological study will be communicated in
due course.
Acknowledgement
Financial support of the Department of Science and Technology (DST), New Delhi (India) under the project entitled
“National Facility for Drug Discovery through New Chemical Entities (NCEs) Development and Instrumentation
Support to Small Manufacturing Pharma Enterprise” is gratefully acknowledged. The authors are also thankful to
Dr. Cecil Kwong for antimycobacterial screening at Tuberculosis Antimicrobial Acquisition and Coordinating
Facility (TAACF), USA.
Supplementary data
Supplementary data associated with this article can be found, in the online version, at doi:
References
12

1.
(a) Phillips, L. Nature 2013, 493, 14; (b) Mitton-Fry, M.J.; Hanna, D. In Neglected Diseases and Drug
Discovery; Palmer, M.J.; Wells, T.N.C. (Eds.) 2012, Chapter 9, pp 228-261.
Ginsberg, A.M.; Spigelman, M. Nat. Med. 2007, 13, 290.
2.
3.
4.
Janin, Y.L. Bioorg. Med. Chem. 2007, 15, 2479 and references cited therein.
World
Health
Organization
(WHO),
global
tuberculosis
control
2010
http://www.who.int/tb/publications/global_report/2010/en/index.html.
5.
6.
7.
8.
Gutierrez-Lugo, M.T.; Bewley, C.A. J. Med. Chem. 2008, 51, 2606.
Nayyar, A.; Jain, R. Curr. Med. Chem. 2005, 12, 1873 and references cited therein.
Rawat, B.; Rawat, D. Med. Res. Rev. 2013, 33, 693 and references cited therein.
Saquib, M.; Gupta, M.K.; Sagar, R.; Prabhakar, Y.S.; Shaw, A.K.; Kumar, R.; Maulik, P.R.; Gaikwad,
A.N.; Sinha, S.; Srivastava, A.K.; Chaturvedi, V.; Srivastava, R.; Srivastava, B.S. J. Med. Chem. 2007, 50,
2942.
9.
10.
11.
Ahirrao, P. Mini Rev. Med. Chem. 2008, 8, 1441 and references cited therein.
Koul, A.; Arnoult, E.; Lounis, N.; Guillemont, J.; Andries, K. Nature 2011, 469, 483.
Blumberg, H.M.; Burman, W.J.; Chaisson, R.E.; Daley, C.L.; Etkind, S.C.; Friedman, L.N.; Fujiwara, P.;
Grzemska, M.; Hopewell, P.C.; Iseman, M.D.; Jasmer, R.M.; Koppaka, V.; Menzies, R.I.; O'Brien, R.J.;
Reves, R.R.; Reichman, L.B.; Simone, P.M.; Starke, J.R.; Vernon, A.A. Am. J. Respir. Crit. Care Med.
2003, 167, 603.
12.
13.
Yew, W.W.; Cynamon, M.; Zhang, Y. Expert Opin. Emerging Drugs 2011, 16, 1.
Ali, M.A.; Samy, J.G.; Manogaran, E.; Sellappan, V.; Hasan, M.Z.; Ahsan, M.J.; Pandian, S.; Yar, M.S.
Bioorg. Med. Chem. Lett. 2009, 19, 7000.
14.
15.
Sunduru, N.; Gupta, L.; Chaturvedi, V.; Dwivedi, R.; Sinha, S.; Chauhan, P.M.S. Eur. J. Med. Chem. 2010,
45, 3335.
Pieroni, M.; Lilienkampf, A.; Wan, B.; Wang, Y.; Franzblau, S.G.; Kozikowski, A.P. J. Med. Chem. 2009,
52, 6287.
16.
17.
18.
19.
20.
21.
Meng, Q.; Luo, H.; Chen, Y.; Wang, T.; Yao, Q. Bioorg. Med. Chem. Lett. 2009, 19, 5372.
Meng, Q.; Luo, H.; Liu, Y.; Li, W.; Zhang, W.; Yao, Q. Bioorg. Med. Chem. Lett. 2009, 19, 2808.
Frieden, T.R.; Sterling, T.R.; Munsiff, S.S.; Watt, C.J.; Dye, C. Lancet 2003, 362, 887.
Wiles, J.A.; Bradbury, B.J.; Pucci, M.J. Expert Opin. Ther. Pat. 2010, 20, 1295.
Minovski, N.; Vracko, M.; Solmajer, T. Mol. Divers. 2011, 15, 417.
Lilienkampf, A.; Mao, J.; Wan, B.; Wang, Y.; Franzblau, S.G.; Kozikowski, A.P. J. Med. Chem. 2009, 52,
2109.
22.
23.
Villemagne, B.; Crauste, C.; Flipo, M.; Baulard, A.R.; Deprez, B.; Willand, N. Eur. J. Med. Chem. 2012,
51, 1.
(a) Kharkar, P.S.; Desai, B.; Gaveria, H.; Varu, B.; Loriya, R.; Naliapara, Y.; Shah, A.; Kulkarni, V.M. J.
Med. Chem. 2002, 45, 4858; (b) Desai, B.; Sureja, D.; Naliapara, Y.; Shah, A.; Saxena, A.K. Bioorg. Med.
Chem. 2001, 9, 1993; (c) Manvar, A.; Pissurlenkar, R.R.S.; Virsodia, V.; Upadhyay, K.; Manvar, D.;
Mishra, A.; Acharya, H.; Parecha, A.; Dholakia, C.; Shah, A.; Coutinho, E.C. Mol. Divers. 2010, 14, 285.
(a) Virsodia, V.; Pissurlenkar, R.R.S.; Manvar, D.; Dholakia, C.; Adlakha, P.; Shah, A.; Coutinho, E.C.
Eur. J. Med. Chem. 2008, 43, 2103; (b) Sarkar, D.; Sarkar, S.; Joshi, R.R.; Khedkar, V.M.; Pissurlenkar,
R.R.; Coutinho, E.C.; Shah, A.K. PCT Int Appl, WO 2011/151701 A1 20111208, 2011; (c) Sarkar, D.;
Sarkar, S.; Joshi, R.R.; Khedkar, V.M.; Pissurlenkar, R.R.; Coutinho, E.C.; Shah, A.K. Indian Pat Appl IN
2010DE01244 A 20120302, 2012.
(a) Manvar, A.; Malde, A.; Verma, J.; Virsodia, V.; Mishra, A.; Upadhyay, K.; Acharya, H.; Coutinho,
E.C.; Shah, A. Eur. J. Med. Chem. 2008, 43, 2395; (b) Upadhyay, K.; Bavishi, A.; Thakrar, S.; Radadiya,
A.; Vala, H.; Parekh, S.; Bhavsar, D.; Savant, M.; Parmar, M.; Adlakha, P.; Shah, A. Bioorg. Med. Chem.
Lett. 2011, 21, 2547; (c) Upadhyay, K.; Manvar, A.; Rawal, K.; Joshi, S.; Trivedi, J.; Chaniyara, R.; Shah,
A. Chem. Biol. Drug Des. 2012, 80, 1003; (d) Virsodia, V.; Shaikh, M.S.; Manvar, A.; Desai, B.; Parecha,
A.; Loriya, R.; Dholariya, K.; Patel, G.; Vora, V.; Upadhyay, K.; Denish, K.; Shah, A.; Coutinho, E.C.
Chem. Biol. Drug Des. 2010, 76, 412.
24.
25.
26.
27.
Manvar A.; Bavishi, A.; Radadiya, A.; Patel, J.; Vora, V.; Dodia, N.; Rawal, K.; Shah, A. Bioorg. Med.
Chem. Lett. 2011, 21, 4728.
Khunt, R.C.; Khedkar, V.M.; Chawda, R.S.; Chauhan, N.A.; Parikh, A.R.; Coutinho, E.C. Bioorg. Med.
Chem. Lett. 2012, 22, 666.
13

28.
For review see: (a) Hari, D.P.; König, B. Angew. Chem. Int. Ed. 2013, 52, 4734 and references cited therein
(b) Murineddu, G.; Loriga, G.; Gavini, E.; Peana, A.T.; Mulé, A.C.; Pinna, G.A. Arch. Pharm. Pharm.
Med. Chem. 2001, 334, 393.
29.
30.
Collins, L.A.; Franzblau, S.G. Antimicrob. Agents Chemother. 1997, 41, 1004.
(a) Strobl, C.; Malley, J.; Tutz, G. Psychol. Methods 2009, 14, 323; (b) Choi, S.-Y.; Shin, J.-H.; Ryu, C.-
K.; Nam, K.-Y.; No, K.-T.; Park Choo, H.-Y. Bioorg. Med. Chem. 2006, 14, 1608; (c) Young, S.S.;
Hawkins, D.M. Methods Mol. Biol. 2004, 275, 317; (d) Rusinko III, A.; Young, S.S.; Drewry, D.H.;
Gerritz, S.W. Comb. Chem. High Throughput Screening 2002, 5, 125; (e) Blower, P.; Fligner, M.;
Verducci, J.; Bjoraker, J. J. Chem. Inf. Comput. Sci. 2002, 42, 393; (f) Zmuidinavicius, D.; Didziapetris, R.;
Japertas, P.; Avdeef, A.; Petrauskas, A. J. Pharm. Sci. 2003, 92, 621; (g) Morshed, M.N.; Cho, Y.S.; Seo,
S.H.; Han, K.-C.; Yang, E.G.; Pae, A.N. Bioorg. Med. Chem. 2011, 19, 907; (h) Chen, X.; Rusinko, A.;
Young, S.S. J. Chem. Inf. Comput. Sci. 1998, 38, 1054; (i) Rusinko, A.; Farmen, M.W.; Lambert, C.G.;
Brown, P.L.; Young, S.S. J. Chem. Inf. Comput. Sci. 1999, 39, 1017; (j) Lu, J.; Zheng, M.; Wang, Y.; Shen,
Q.; Luo, X.; Jiang, H.; Chen, K. J. Comput.-Aided Mol. Des. 2011, 25, 885; (k) Horvath, D. SAR QSAR
Environ. Res. 2001, 12, 181; (l) Burton, J.; Ijjaali, I.; Barberan, O.; Petitet, F.; Vercauteren, D.P.; Michel,
A. J. Med. Chem. 2006, 49, 6231; (m) Deconinck, E.; Zhang, M.H.; Coomans, D.; Vander Heyden, Y. J.
Chem. Inf. Model 2006, 46, 1410.
31.
32.
33.
Accelrys, I. Cerius2. San Diego, CA, USA; 1998.
See, Table 1S in supporting informations.
For activity prediction, see Table 2S in supporting informations.
14

Graphical abstract
Synthesis And Binary QSAR Study of Antitubercular
Quinolylhydrazides
Atul Manvar, Vijay Khedkar, Jignesh Patel, Vipul Vora, Narsinh Dodia, Gautam Patel,
Evans Coutinho, and Anamik Shah