CFiapter - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/14408/13/13_chapter 6.p… · compounds were first of their class to show this kind of activity . . Here, establishing

CFiapter 6

~jI{]{Stutfies on Punctiona{izeajl{~nolS ana C-3 a{~{j aryl4~C-2,3-aitfeoX}

. :J{e~nopyranositfes as Jlntitu6ercu14r Jleents

Functionalized Alkenols & 2,3-Dideoxy Hexenopyranosides as Antitubercular Agents

6.1 INTRODUCTION AND BASIS OF WORK Mycobacterium tuberculosis, the leading causative agent of tuberculosis (TB), is

responsible for the morbidity and mortality of a large population worldwide. Moreover,

the opportunistic TB infections due to AIDS and emergence of drug resistant TB strains

have made its chemotherapeutic strategies increasingly ineffective. Traditionally, it has

relied heavily on a limited number of drugs such as isonicotinicacid hydrazide,

rifampicin, ethambutal, streptomycin, ethionamide, pyrazinamide, fluoroquinolones,

etc. I,2 However, many of these drugs have different disadvantages such as prolonged

treatment schedules, host toxicity, ineffectiveness against resistant strains etc. Thi$ has

motivated for the search of new chemical prototypes capable of rapid mycobactericidal

action with shortened duration of therapy, reduced toxicity and enhanced activity against

drug resistant strains and also against the latent bacteria for its control.3-7 This paradigm

shift in the strategies has drawn attention towards the outer cell wall and membranes of

the pathogen as an attractive target for exploring new drugs.8,9

In mycobacteria, the cell wall structure consists of a dense network of cross

linked sugar residues esterified with mycolic acid at the ends.IO,II This understanding has

prompted for the investigation of various sugar prototypes with distinct characteristics as

potential antitubercular agents.12-IS In. recent literature, some 2H- pyran-3(6~ne

derivatives are reported to exhibit significant activity against gram-positive bacteria. 16 A

variety of carbohydrate derivatives are also known to interfere with the cell wall

biosynthesis of M tuberculosis.13,14 These studies suggested that arabinogalactan linked

disaccharides and decaprenolphosphoarabinosides interfere with the mycobacterial

arabinosyltransferases. Moreover, some iminosugars are inhibitors of the UDPGalf

transferase (Uridine diphosphate galactofuranosyl transferase), in mycobacterial cell wall

biosynthesis. IS All these studies emphasize on modified carbohydrates as potential

scaffolds to interfere with the components of mycobacterial cell wall biosynthesis.

In this background, the 'functionalized heptenol17 and octenolI8 derivatives

(functionalized alkenols, acyclic sugars, Fig. la)' and 'C-3 alkyl! arylalkyl-2,3-dideoxy

hexenopyranoside derivatives 19 (cyclic sugars, Fig. I b)' have been developed in our

laboratory as a part of our laboratory's ongoing new drug development program on the

design of antitubercular agents. These sugar templates are considered as a potential

source of new molecular scaffolds with strategically positioned functional groups that

140


will selectively interact and communicate with the complementary groups / sites of the

pathogen's cell-wall structure and receptor(s) therein. To the best of our knowledge, these

compounds were first of their class to show this kind of activity .

. Here, establishing a correlation between the structure and the associated activity

will come to aid in furthering the understanding of the system under investigation. In

these studies parameterization of chemical structure plays a pivotal role. In enumeration

of chemical structures, it is important to note that in isolation a data point is only a

qualified number. A collection of such qualified numbers makes a variable or descriptor.

In mathematical models each and every variable communicate with all other variables. A

meaningful inter- and intra-variables communications result in the evolution of models

with predictive value. With this philosophy quantitative structure-activity relationship

(QSAR) study have been conteinplated on functionalized heptenol and octenol

derivatives (functionalized alkenols, (Fig. la) and C-3 alkyl / arylalkyl-2,3-dideoxy

hexenopyranosides (Fig. lb) using different types of molecular descriptors obtained from·

. Dragon20 software. The combinatorial protocol in multiple linear regression(CP-MLR)

approach,21 have been used for model development. The detaifed procedural aspects and

results are discussed here.

z

a b

Fig. 1. (a) General structure of the functionalized alkenols associated with antitubercular activity. These compounds were derived starting from glucal or galactal derivatives by opening the respective pyranose rings at the anomeric carbon (C1). For the purpose of correspondence with the parent sugars, in this structural template the carbon corresponding to the anomeric carbon of pyranose has been opted as '1' (that is C1)and the succeeding carbons as '2' (C2), '3' (C3) etc. Here, for the structural template corresponding to nitro/acetamido alkenol derivatives (Table 2), Rs and ~ is OAc, and R7 is CH20Ac and for that of chloro/amino alkenol derivatives (Table 3), Rs is OBn and C3-C4 centers are joined with double bond. (b) Substituted-2,3-dideoxy hexenopyranosides RI and R2 are alkyl and aryl groups respectively.

141


6.2 QSAR STUDY ON FUNCTIONALIZED ALKENOLS For functionalized alkenol (heptenol and octenol) derivatives different empirical,

constitutional and topological descriptors of the compounds have been opted from

Dragon software.20 The descriptor classes considered in the study along with their

definitions and scopes in addressing the molecular structure have been presented in Table

1. With all these descriptors as independent parameters, the structure-activity relations of

the antitubercular activity of the functionalized alkenol derivatives (Tables 2 and 3) have

been discovered using the CP-MLR approach.21 As the total number of descriptors

involved in this study are about two hundred and ninety for each set of compounds, only

the names of descriptor classes and participating descriptor names have been addressed in

the discussion. The details of computional procedure is discussed in the methodology.

6.2.1 MATERIALS AND METHODS

6.2.1.1 Dataset. Two series of compounds, nitro/ acetamido derivatives (functionalized

heptenols, Table 2)17 and chloro/ amino alkenol derivatives (functionalized octenols,

Table 3)18 have been taken up in this study along with their antitubercular activity as

logarithm of the inverse of minimum inhibitory concentration (-logMIC where MIC in

moles per liter against M tuberculosis, H37Rv). As these compounds (Fig. la) have been

synthesized starting from either glucal or galactal derivatives by opening the respective

pyranose ring at the anomeric carbon, their main backbone has at least six carbons in

them. To maintain correspondence with the parent templates, the numbering of main

carbon backbone of these compounds (Fig. la) has been done according to the numbering

convention of pyranose sugar moiety (anomeric carbon as CI, and so on). For the

computation of molecular descriptors, the structures of the compounds have been

generated using the following protocol. As the compounds 12, 13, 20 and 21 (Tables 2

and 3) are some of less substituted .ones among the compounds of the study, they have

been considered as the core templates (core compounds) for rest of the analogues. The

2D ChemDraw structure of core template has been converted into a 3D-object using the

default conversion procedure implemented in the CS Chem3D Ultra?2 The so generated

3D-conformer of the compound has been subjected to energy minimization in the

MOPAC module, using the AMI procedure for closed shell systems, implemented in the

CS Chem3D Ultra.22

142 .

FunctionalizedAlkenols & 2,3-Dideoxy Hexenopyranosides as Antitubercular Agents

Table 1. The descriptors classes used for the analysis of antitubercular activity of nitro!

acetamido alkenol (Table 2) and chloro! amino alkenol (Table 3) derivatives and number

of models identified in each class

Descriptor Class

(acronymt

Empirical (EMP)

Constitutional

(CONST)

.. Definition and scope

(Number of descriptors)

Descriptors represent the counts of non

single bonds, hydrophilic groups and ratio of

the number of aromatic bonds and total

bonds in H-depleted molecule (2)

Dimensionless descriptors, independent from

molecular connectivity and conformations

(24)

Descriptors per

model

(Number of models)h

Table 2 Table 3

Topological (TOPO) Descriptors obtained from molecular graphs 1(1);

and independent of conformations (68) 2(3)

Molecular Walk

counts (MWC)

Modified Burden

eigenvalues (BCUT)

Galvez Topological

charge indices

(GVZ)

Descriptors representing self returning walks

counts of different lengths (13)

. Descriptors representing positive and

negative eigen values of the adjacency

matrix, weights the diagonal elements & atoms (64)

Descriptors representing the first 10 eigen

values of corrected adjacency matrix (21)

2D autocorrelations Molecular descriptors calculated from the

(2DAUTO) molecular graphs by summing the products

a, Reference 20.

of atom weights of the terminal atoms of aU the paths of the considered path length (the

lag). (96)

1(3);

2(20)

2(5)

b, Models emerged from CP-MLR protocol with filter-l as 0.3; filter-2 as 2.0; filter-3 as 0.74; filter-4 as 0.3 s.Q2 S.1.0, number of compounds in each dataset are eleven.

143


Table 2. Observed and modeled antitubercular activity of nitro/ acetamido alkenols (Fig. la)

Compd.a Z R. Rz Cs* B3,4 Ii -logMIC

Obsd.c Eq.1Q Eq.2Q

Eq.3i1 Eq.4i1 Eq.Sil

N02 H OH S db na . 4.15 4.17 3.91 3.97 4.07

2 N02 H OH R db 4.12 4.15 4.17 3.91 3.97 4.07

3 N02 CH3 OH S db 3.84 3.79 3.89 3.80 3.82 3.74

4 N02 H H S db na 3.82 3.62 3.67 3.69 3.68

5 N02 H H R db 3.50 3.82 3.62 3.67 3.69 3.68.

6 N02 CH3 H S db na 3.42 3.39 3;56 3.55 3.43

7 N02 CH3 H R db 3.52 3.42 3.39 3.56 3.55 3.43

8 NHAc H OAc S sb na 3.71 2.69 4.68 4.77 2.54

9 NHAc II OAc R sb na 3;71 2.69 4.68 4.77 2.54

10 NHAc CH3 OAc S sb na 3.37 1.88 . 4.57 4.63· 1.53

11 NHAc CH3 OAc R sb na 3.37 1.88 4.57 4.63 1.53

12 NHAc· H H S sb 4.72 4.73 4.70 4.80 4.83 4.70

13 . NHAc H H R sb 4.72 4.73 4.70 4.80 4.83 4.70

14 ·NHAc CH3 H S sb 4.44 4.43 4.43 4.29 4.29 4.33

15 NHAc CH3 H R sb 4.14 4.43 4.43 4.29 4.29 4.33

16 N02 H H S db 4.70 4.47 4.49 4.48 4.46 4.66

17 N02 H H R db na 4.47 4.49 4.48 4.46 4.66

18 N02 CH3 H S db 4.72 4.14 4.22 4.32 4.27 4.31

19 N02 CH3 H R db 3.82 4.14 4.22 4.32 4.27 4.31

a, In compounds 4 t07, Cl and C2 are connected with double bond.

b, Bond order between C3 and C4, 'sb' for single bond and 'db' for double bond.

c, Reference 17.

d, Activity from corresponding equation.

144


Table 3. Observed and modeled antitubercular activity of chlorol amino alkenols (Fig.

la)

Compd.8 Z RJ R2 ~ R7 Cs* -logMIC

Obsd.b Eq.5c Eq.6 c Eq.7 c

20 CH2CI Ac H OBn CH20Bn S na 4.84 4.74 5.07

21 CH2CI Ac H OAc CH20Bn R 5.16 4.94 4.89 5.01

22 CH2CI Ac H OAc H S 4.13 4.32 4.49 4.29

23 CH2N(CH2)sd Ac H OAc CH20Bn R 4.00 4.35 4.25 4.31

24 CH2N(CH2)4 e Ac H OAc CH20Bn R 4.29 4.48 4.46 4.31

25 CH2N(C2HSh Ac H OAc CH20Bn R 3.99 3.80 3.92 4.31

26 CH2N(C2~)2NMef Ac H OAc CH20Bn R 4.32 4.33 4.37 4.31

27 CH2N(CH2)Sd Ac H OAc H S 3.89 3.85 3.83 3.79

28 =CH2 CN OH OBn CH20Bn S 3.97 4.10 4.03 4.05

29 =CH2 CN OH OH CH20Bn R 3.88 3.86 3.79 3~80

30

31

CH2N(C2~)2NMef CN OH OBn CH20Bn S

CH2N(C2~)2NMef CN OH OH CH20Bn R'

a, In compounds 20 to 27 Cl and C2 are connected with double bond. b, Reference 18. c, Activity from corresponding equation. d, Piperidinyl methyl. e, Pyrrolidinyl methyl. f, N-methyl-piperazinyl methyl.

4.66 4.35 4.34 4.38

4.28 4.20 4.21 4.04

To maintain a well-defined conformer relationship across the compounds of the

study, the structures of all other compounds have been generated, by appending

appropriate changes to the core template followed by 3D conversion and energy

minimization as implemented incase of core compounds. In this study, structural models

of compounds 1, 3, 4, 6, 8, 10, 14, 16, 18 (Table 2), 2, 5, 7, 9, 11, 15, 17, 19 (Table 2),

22,27,28, 30 (Table 3) and 23-26, 29, 31 (Table 3) have emerged from the core template.

of compounds 12, 13,20 and 21, respectively. In the structure building process of all the

compounds, the disposition of all functional groups at the asymmetric centers (except the

ones cO.rresponding to C 1 and C2) and at the double bounds, where applicable, have been

maintained according to our synthetic report. 17,18 In the absence of any information on the

preferences of configurations of C I and C2, they have been arbitrarily kept as Sand R,

respectively. This has no qualitative or quantitative effect on the parameters of the dataset

of the compounds, as these structures have been used only to generate 2D- descriptors.

145


All these Chem3D Ultra generated structures of the compounds have been converted into

Tripos MOL2 format (.ML2 files) in HyperChem23 to enable them to be portable to

Dragon software20 for computing the various 2D- molecular descriptors. The CP-MLR

computational procedure21 and model validation are briefly described below.

6.2.1.2 Computational procedure and model development. The CP-MLR is a

'filter' based variable selection procedure for model development.21 Its procedural

aspects and implementation are discussed in Chapter 3. The thrust of this procedure is in

its embedded 'filters'. They are briefly as follows: filter-I seeds the variables by way of

limiting inter-parameter correlations to predefined level (default acceptable value ::;0.3); .

filter-2 controls the variables entry to a regression equation through I-values of

coefficients (default acceptable value 2: 2.0); filter-3 provides comparability of equations

with different number of variable in terms of square-root of adjusted mUltiple correlation

coefficient of regression equation, r-bar (default acceptable value 2: 0.74 ); filter-4

estimates the consistency of the equation in terms of cross-validated R2 or Q2 with leave

one-out (LOO) cross-validation as default option (default acceptable limits are 0.3 ::; Q2::; .

1.0).

Throughout this study, for the filters-I, 2, 3 and 4 of CP-MLR the thresholds were

assigned as· 0.3, 2.0, 0.74 and 0.3 ::; Q2 ::; l.0, respectively. They make the variable

selection process efficient and lead to unique solutions .. Furthermore, to discover any

chance correlations associated with the models identified in CP-MLR, each cross

validated model has been subjected to randomization test by repeated randomization of

the biological response?4,25 The datasets with randomized response vector have been

reassessed by multiple regression analysis. The emerging regression equations, if any,

with correlation coefficients better than or equal to the one corresponding to unscrambled

response data were counted. Every model has been subjected to one hundred such

simulation runs. This has been used as a measure to express the percent chance

correlation of the model under scrutiny. The CP-MLR protocol has been applied with

default filter thresholds on the two dataset~ to identify the all-possible models that could

emerge from the descriptors of each set of compounds.

146


6.2.2 RESULTS AND DISCUSSION

In multi-descriptor class environment, exploring for models along the descriptor

class provides an opportunity to understand the phenomenon under investigation vis-a.-vis

the concept embedded in the descriptor. With this, attempts have been made to develop

one- and two-descriptor models for the antitubercular activity of the nitro/ acetamido

alkenol derivatives (Table 2) and chloro/ amino alkenol derivatives (Table 3). For nitro/

acetamido alkenol derivatives (Table 2), the study has led to four models in the

topological (TOPO) class descriptors and twenty-three models in the 2D autocorrelations

(2DAUTO) class descriptors (Table 1). In case of chlorol amino alkenol derivatives

(Table 3), only the 2DAUTO class descriptors have resulted in five models (Table I).

The essence of all the models obtained from the study has been provided in Tables 4 and

6 in the form of identified descriptor's average of regression coefficient along with its

standard deviation across the models and the total incidence corresponding to all the

models. This, while providing th~ averages of the estimated regression coefficients of all

the identified descriptors, shows their variance across the models emerged from the study

as well. To maintain brevity, the complete regression equations have been shown for

selected models only. The following regression equation represents a. TOPO class

structure-activity model of the nitro/ acetamido alkenol derivatives (Table 2).·

-logMIC = - 0.034(0.0 16)TWC - 0.824(0. 192)PCR + 9.691

n=ll, r=0.840, Q2=0.487, s=0.296, F=9.S7. (1)

In this and all other regression equations, n is the number of compounds, r is the

correlation coefficient, Q2 is cross-validated R2 from leave-one-out (LOO) procedure, s is

the standard error of the estimate and F is the F-ratio between the variances of calculated

and observed activities. The values given in the parentheses are the standard errors of the

regression coefficients. Also, in the randomization study (hundred simulations per model)

f\one of the identified models has shown any chance correlation. The four models,

including the above-presented one, emerged from TOPO class have shared five

descriptor among themselves (Table 4). In Eq. 1, the activity is correlated with ratio of

multiple path counts to path counts (PCR) alone to the order of 0.74. In addition to this,

activity is correlated with one more TOPO descriptor that is total multiple walk count

(TPCM; r=0.79).

147


Table 4. Descriptors identified in modeling the antimicobacterial activity of nitro/ acetamido

alkenol derivatives (Table 2) along with their average regression coefficients

Descriptor" Av reg coef (sd)Total incidence6 Descriptor" Av reg coef (sd)Total incidence6

Individual Combined Individual Combined. EMP 2DAUTO Vi -1.918(0.170)2 ATS3m -7.584(-)1 Hy -0.561(0.080)2 ATSlv -46.804(-) 1

ATS3v -27.444(0.670)2 CONST ATS7v -32.081(-)1 Mv -31.900(-) I ARS8v -31.892(8.200)2 nDB -0.4 79(0.040)2 ATS2p -67.242(-) I nOH -0.326(0.050)4 ATS3p -36.461(0.582)2 nHD -0.294(0.026)4 ATS4p -13.241(-)1

ATS7p -43.365(-)1 TOPO ATS8p -21.634(3.031 )2 TWC -0.034(-)1 MATSlm 17.082(0.833)4 TPCM -0.003(-)1 MATS3m 6.863(1.361)2 PCR -0.802(0.0285)3 MATS6m -6.116(0.929)4 PCD -0.244(0.016)2 MATS7m -17.016(1.092)2 SMTIV -0.0004(-)1 MATS Iv -12.547(2.707)3 GSI -0.162(-) 1 MATS4v 7.761(-)1

MATS3e 7.743(-)1 MWC MATSlp -8.622(-)1 MWC04 -0.195(-)1 GATS6m 3.206(-)1

GATS7m 5.959(0.502)3 BCUT GATS Iv . 3.536(0.007)2 BEPm2 -11.317(0.015)2 GATS2v 4.697(-)1 BEPv2 -6.728( 1.346)4 GATSle -1.653(0.023)2 BEPe2 -"0.858(1.257)3 GATS2e. -4.245(-)1 BEPp2 -6.225(0.099)2 GATSlp 3.131(0.007)2

GATS2p 4.577(0.604)2 GVZ GATS3p -3.256(-)1 GGI3 -1.357(-)1 JGI3 -35.204(-) 1

a, The descriptors are identified from the one and two parameter models emerged from CP-MLR

protocol with filter-I as· 0.3; filter-2 as 2.0; filter-3 as 0.74;filter-4 as 0.3~Q2~1.0, number of

compounds in the study are eleven; EMP: Vi, unsaturation index; Hy, hydrophilic factor; CONST:

. Mv, mean atomic van der Walls volume; nOB, number of double bonds; nOH, number of hydroxyl

groups; nHD, number of donor atoms for H-bond; nCI, number of chloro groups present in the

molecule; TOPO: TWC, total walk count; TPCM, total multiple walk count; PCR, ratio of multiple

path counts to path ·counts; PCD, difference of multiple path counts to path counts; SMTIV, Schultz

moleCUlar topological index by valence vertex degrees; GSI, Gordon-Scantlebury index; MWC:

MWC04 is molecular walk count of order 4; BCUT: in BEPwn, 'BEP' represents Positive Burden Eigenvalue; penultimate character 'w' indicates the weighting property used in the computation which

may be either 'm' Le. atomic masses or 'v' Le. atomic van der Walls volumes or 'e' i.e. Sanderson

, electronegativities or 'p' i.e. atomic polarizabilities; the last character 'n' indicates the eigen value

level; GVZ: GGI3, topological charge index of order 3; JGB, mean topological charge index of order

3; 2DAUTO: In the descriptors of autocorrelation of topological structure of A TSnw (Broto-Moreau),

MATSnw (Moran) and GATSnw (Geary) the penultimate character, 'n' indicates the autocorrelation

148


vector of lag n corresponding to the number of edges in the fragment unit considered in the

computation and the last character 'w' refers to the same as in case ofBCUT. Also see reference 20. b, The average regression coefficient of the descriptor corresponding to all models, its standard

deviation (s.d.) and the total number of its incidence. The arithmetic sign of the coefficient represents

the actual sign of the regression coefficient in the models. A blank in 'Individual' column indicates

that descriptor has not participated in any modeL The 'Combined' column corresponds to the dataset

of all descriptor classes together and shows only those descriptors which additional to the 'individual'

column.

Also, all the participating TOPO class descriptors' regression coefficients have

been associated with negative sign (Table 4). Among the descriptors, PCR has the highest

incidence of three. (Table 4). The emerged models suggest that a smaller PCR value

would be favorable for the activity, which in other words favors more saturated system

for better activity. The negative regression coefficient of TPCM, which accounts for the

multiple bonded centers in the system, also suggests that compounds with less multiple

bonds are better for the activity. The regression coefficients of total walk count (TWC)

and Gordon-Scantlebury index (GSI) suggest that a compound with branching will be

less favorable when compared to its linear homologue. Here, Schultz molecular

. topological index by valence vertex degrees (SMTIV) is an index that addresses both

branching and multiple bonded centers in a given compound. In homologous series of

compounds, the multiple bonded centers markedly increase the value of SMTIV whereas

branching effect decreases the same but relatively to a lesser extent. This in tum suggests

that the unsaturation in the compounds is more detrimental to the activity when compared

to the branching effect. All these descriptors collectively indicate a preference for

compounds with more saturated and less branched structural templates for antitubercular

activity.

The 2DAUTO class descriptors also represent the topological structure of the

compounds. But these descriptors are more complex in nature when compared to the.

TOPO class descriptors. The 2DAUTO descriptors considered in the study have their

origin in autocorrelation of topological structure of Broto-Moreau (ATS), of Moran

(MATS) and of Geary (GATS)?O,26 The computation of these descriptors involve the

summations of different autocorrelation functions corresponding to the different fragment

lengths and lead to different autocorrelation vectors corresponding to the lengths of the

structural fragments?6 Also a weighting component in terms of a physicochemical

149


property has been embedded in this descriptor. As a result these descriptors address the

topology of the structure in association with a selected physicochemical property. In the

descriptor's nomenclature, the penultimate character, a number, indicates the number of

consecutively connected edges considered in its computation and is called as the

autocorrelation vector of lag n (corresponding to the number of edges in the unit

fragment). The very last character of the descriptor's nomenclature indicates the

physicochemical property considered in the weighting component - m for mass or v for

volume or e for Sanderson electronegativity or p for polarizability - for its computation.

The philosophy embedded in these descriptors and their computational aspects are

available in different sources?0,26 The twenty-three models emerged from this descriptor

class have shared twenty-three descriptors among themselves (Table 4). The following

three regression equations (Eqs. 2 - 4) represent the selected models of this class of

descriptors;

-logMIC = -19.49i(9.572)ATS8p -16.243(4.004)MATS7m + 10.780

n=ll, i=0.869, Q2=0.604, ' s=0.270, F=12.31 (2)

-logMIC =- 5.324(2.589)MATS6m + 3.541(0.797)GATSlv - 6.492

n=11, r=0.865, s=0.273, . F=11.92 (3)

-logMIC = - 5.326(2.605)MATS6m + 3.136(0.712)GATSlp - 5.855

n=11, r=0.863, s=0.275, F=11.72 (4)

In all the three models, the activity is correlated to a significant extent (r=0.73 to

0.79) with one descriptor, which is either Moran autocorrelation lag seven weighed by

atomic masses (MATS7m) or Geary autocorrelation lag one weighed by van der WallsWqc'/~ . . ~

volumes (GATS 1 v) or Geary autocorrelation lag one weighed by atomic polarizabilities

(GATSlp). PCR, TPCM of TOPO class and these three descriptors are highly

intercorrelated with each other (r=0.90 to 0.99). Furthermore, the trend of the formed

models and descriptors participated therein (Table 4) indicate that descriptors of lag one

have accounted for the maximum number of models (cumulative frequency is fifteen)

followed by lag three (cumulative frequency is eight) and lag seven (cumulative.

frequency is six) descriptors. A study of the correlation matrix of these descriptors

150


revealed that the autocorrelation vectors of each lag formed by mass, volume and

polarizability weightings are highly intercorrelated.

Also, in addition to the three descriptors (MATS7m, GATSlv and GATSlp) of

Eqs. 2 - 4, seven more descriptors (ATS3v, ATS7v, ATS2p, ATS3p, MATSlp and

GATS7m) have shown correlation individually with the activity to the order of 0.70 to

0.79. In this class, the lag one. descriptors are simplest among all. They do not

differentiate between normal and branched structural templates, but identify the loops and

multipully connected vertices in them; These descriptors also suggest that unsaturation in

the structural templates is not favorable for the activity. The participation of descriptors

of lag six and seven may be viewed in terms of association of activity information

content with the six and seven. centered structural fragments. Most. of the identified

descriptors have mass, volume and! or polarizability ~ physicochemical weighting

component in them indicating their influence on the activity. However, further

deciphering of the information content of these descriptors is very complex as their

computations involve integration of the structural fragments and due to this it is not

possible to traverse backward from a higher state to a lower one.26 In Eqs. 1 to 4 and for

all other identified models the rand Q2 values are not of very high order. It is mainly due

to the deviations in the calculated as well as predicted activities·of compounds 18 and 19.

Exclusion of these two compounds from the dataset improved the Eqs. 1 to 4 as shown in

Eqs. la to 4a, respectively. The participating descriptors correspond to these and other

equations for nitro! acetamido alkenol derivatives (Table 2) are shown in Table 5.

-logMIC = -':0.035(0.011)TWC - 0.834(0.132)PCR + 9.797

n=9, r=0.935, s=0.203, F=20.725 (la)

-logMIC = - 21.071(6.443)ATS8p -15.806(2.566)MATS7m + 1 L281

n~9, r=0.957, s=0.166, F=32.58 (2a)

-logMIC = - 5.477(1.690)MATS6m + 3.554(0.511)GATSlv -6.523

n=9, r=0.952, s=0.175, F=29.06 (3a)

-logMIC =- 5.307(1.752)MATS6m + 3. 136(0.469)GATSIp - 5.858

n=9, r=0.948, Q2=0.763, s=0.181, F=26.83 (4a)

151


Table 5. Descriptors correspond to all equations (Eq. 1 to 4 and Eq.8) for nitro/ acetamido alkenol derivatives (Table 2)

Com pd. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

TWca 139.67 139.67 149.95 130.58 130.58 140.09 140.09 165.03 165.03 175.07 175.07 135.97 135.97 145.02 145.02 130.58

.130.58 140.09 140.09

0.891 0.891 0.897 1.672 1.672 1.761 1.761 0.37 0.37

0.357 0.357 0.341 0.341 0.327 0.327 0.881 0.881 0.891 0.891

ATS8pa

0.315 0.315

0.33 0.33 0.33'

0.346 0.346 0.336 0.336 0.345 0.345 0.321 0.321 0.334 0.334 0.321 0.321 0.336 0.336

MATS6ma

-0.006 -0.006 0.017

-0.018 -0.018 0.008 0.008

-0.012 -0.012 0.002 0.002

-0.041 -0.041 0.048 0.048

-0.046 -0.046

-0.02 . -0.02

MATS7ma

0.029 0.029 0.028 0.045 0.045

0.04 0.04

0.095 0.095 0.134 0.134

-0.011 -0.011

-0.01 -0.01 0.002 0.002 0.001 0.001

GATS Iva 2.93 2.93

2.933 2.843 2.843 2.851 2.851 3.137 3.137 3.127 3.127 3.128 3.128 3.118 3.118

3.03 3.03

3.024 3.024

GATSlpa 3.122 3.122 3.113 3.015 3.015 3.013 3.013 3.367 3.367 3.346 3.346 3.339 3.339 3.317 3.317 3.21 3.21

3.194 3.194

BEPvr 3.343 3.343 3.399 3.357 3.357 3.412 3.412 3.389 3.389 3.432 3.432 3.363 3.363 3.419 3.419 3.331 3.331

3.39 3.39

a, TWC, total walk count; PCR, ratio of multiple path counts to path counts; ATS8p, Broto

Moreau autocorrelation descriptor of lag 8 weight by atomic polarizabilities; MA TS6m and MATS7m are Moran autocorrelation descriptor of lag 6 and 7 respectively weight by atomic

masses; GATSlv and GATSlp are Geary autocorrelation descriptor of lag 1 weight by atomic van der Walls volumes and atomic polarizabilities respectively; BEPv2 Positive Burden Eigenvalue of level 2 weight by atomic van der Walls volume.

However, except the deviations in the calculated! predicted activities we have no

other reason to exclude compounds 18 and 19 from the analysis. The regression

. coefficients of all the descriptors in Eqs. la to 4a are in general agreement with those of

the Eqs. 1 to 4 as well as with those presented in Table 4.

The five models of chlorol amino alkenol derivatives (Table 3) have shared six

2DAUTO class descriptors among themselves (Table 6). The following equations (Eqs. 5

and 6) represent the selected structure-activity models of these compounds.

-logMIC = 11.002(3.571)ATS7v - 3.509(0.892)GATS4m + 7.068

n=11, r=0.856, Q2=0.443, s=0.224, F=1O.95 (5)

-logMIC = 6.538(2.243)ATS8in - 2.512(0.931 )GATS4m + 6.996

n=ll, r=0.846, s=0.230, F=1O.09 (6)

152


Table 6: Descriptors identified in modeling the antimicobacterial activity of chlorol amino alkanol derivatives (Table 3) along with their average regression coefficients

Descriptora Av reg coer (sd) Total incidence Individual Combined

CONST nel BCUT BEPml BEPm7 2DAUTO ATS8m ATS3p ATS7p ATS3v ATS7v GATS4m

6.538(-)1 13.778(-)1 11.090(-)1 13.778(-)1 11.002(-)1

-3.381(0.508)5

0.702(-)1

26.200(-)1 2.144(0.000)2

a, The descriptors are identified from the one and two parameter models emerged from CP-MLR

protocol with filter-l as 0.3; filter-2 as 2.0; filter-3 as 0.74; filter-4 as 0.3:SQ2:sl.O, number of

compounds in the study are eleven; CONST: nel, number of chloro groups; HCUT: in BEPmn,

'BEP' represents Positive Burden Eigenvalue, penultimate character Om' indicates the atomic

masses used in the computation, the last character On' indicates the eigenvalue level; 2DAUTO: In the descriptors of autocorrelation of topological structure of ATSnw (Broto-Moreau) and

GATSnw (Geary) the penultimate character, On' indicates the autocorrelation vector of lag n

corresponding to the number of edges in the fragment unit and Ow' indicates the weighting

property which may be either Om' i.e. atomic masses or 'p' i.e. atomic polarizabilities or 'v' i.e.

atomic van der Walls volumes. Reference 20. b, The average regression coefficient of the descriptor corresponding to all models, its standard

deviation (s.d.) and the total number of its incidence. The arithmetic sign of the coefficient

represents the actual sign of the regression coefficient in the models. A blank in 'Individual'

column indicates that descriptor has not participated in any model. The 'Combined' column

corresponds to the dataset of all descriptor classes together and shows only those descriptors

which additional to the 'individual' column.

The participating descriptors correspond to these equations (Eqs. 5 and 6) are

shown in Table 7. For all the models emerged here, Geary autocorrelation lag four

weighed by atomic masses (GATS4m) is one common descriptor. The remaining five

descriptors have come from Broto-,Moreau autocorrelation of topological structure (ATS)

with mass (m) or volume (v) or polarizability (P) as the weighting parameter (Table 6).

All these five ATS descriptor are highly intercorrelated (r=0.86 to 0.99). The correlation

of activity with autocorrelation vectors of lags three, four and seven indicate that

structural fragment of corresponding lengths are enriched with activity information.

153


Table 7. The participating descriptors correspond to Eqs. 5 to 7 of chloro/ amino alkanol

derivatives (Table 3)

Compd. ATS7va

20. 0.417 21. 0.418 22. 0.388 23.· 0.382 24. 0.377 25. 0.371 26. 0.376 27.0.347 28. 0.41 29. 0.391 30. 0.383 31. 0.362

0.4 0.414 0.385 0.338 0.349 0.335 0.35

0.286 0.366 0.331 0.353 0.324

GATS4ma

1.941 1.917 1.999 1.972

1.92 2.095 1.958 2.004 2.132 2.139 1.976 1.951

1 1 1 o o o o O. o o o o

3.176 3.145 2.811 3.147 3.147 3.146 3.147 2.907 3.029 2.909 3.182 3.022

a ,A TS7v and ATS8m are Broto-Moreau autocorrelation descriptor of lag 7 and 8 weight by atomic van der Walls volumes and atomic mass respectively; GATS4m, Geary autocorrelation descriptor of lag 4·weight by atomic masses; n_CL, number of chloro groups; BEPm7, Positive Burden Eigenvalue of level 7 weight by atomic mass.

At the same time these descriptors indicate the role of physicochemical properties

such as mass, volume and! or polarizability of the compounds in deciding the activity. In

this dataset, exclusion of compound 23 further improved the quality of the models as

shown by Eqs. 5a and 6a, which correspond to Eqs. 5 and 6 respectively.

-logMIC = 11.098(3.112)ATS7v - 3.716(0.785)GATS4m + 7.482

n=lO, r=0.903, Q2=O.588, s=0.195, F=15.41 (5a)

-logMIC = 6.179(2.212)ATS8m-'- 2.693(0.922)GATS4m + 7.510

n=lO, r=0.868, Q2=0.470, s=0.225, F=1O.70 (6a)

As the order and magnitude of the regression coefficients have remained the

same, the equations formed by exclusion of compound(s) also convey the same

information as their predecessors.

The structure-activity relations of the antitubercular activity of the nitro/

acetamido alkenol derivatives (Table 2) and chloro/ amino alkenol derivatives (Table 3)

have been further investigated by creating two more datasets (datasets 'A' and 'B') for

154


each series of compounds. Of these, dataset 'A' corresponds to all descriptor classes,

which have not yielded any model individually. This has been chosen to investigate the

collective information content of left out descriptor classes vis-a.-vis the activity of the

compounds. And the other, dataset' B', contains all the descriptor classes that is selected

as well as left out classes of the study. Under the study conditions, the dataSet 'A' of

nitro/ acetamido alkenol derivatives (Table 2) did not result in any model. However, the

dataset 'A' corresponding to chloro/ amino alkenol derivatives (Table 3) has resulted in

only one two-descriptor model (Eq. 7). Eq. 7a is a variant of Eq. 7 formed by excluding

compound 23 from the dataset.

-logMIC = O. 702(0. 182)nCI + 2. 144(0.578)BEPm7 - 2.439

n=ll, r=0.857, Q2=0.3IS, s=0.223 , F=11.03

-logMIC = 0.679(0. 1 67)nCI + 2.327(0.540)BEPm7 - 2.962'

n=10, r=0.893, Q2=0.4S0, s=0.203, F=13.S4

(7)

(7a)

This clearly indicates that the descriptor classes other than TOPO and/or

2DAUTO classes have insufficient or marginal information (corresponding to the

activity) within themselves - independently as well as collectively - to come up to the

level of model formation. However, under the same conditions, dataset 'B' of nitro/

acetamido alkenol derivatives (Table 2) and chloro/ amino alkenol derivatives (Table 3)

have resulted in sixty-one and seven models, respectively. In dataset 'B', the models

. obtained over and above the TOPO and/or 2DAUTO individual classes indicate the

cooperative participation and contribution of dataset 'A' in combination with the TOPO

and/or 2DAUTO classes. All the descriptors of the additional models emerged from the

dataset 'B' of the compounds of Tables 2 and 3 have been listed in Tables 4 and 6,

respectively.

In nitro/ acetamido alkenol derivatives (Table 2), the empirical (EMP) (two

descriptors), constitutional (CONST) (four descriptors), Molecular Walk counts (MWC)

(one descriptor), Modified Burden eigenvalues (BCUT) (four descriptors) and Galvez

Topological charge indices (GVZ) (two descriptors) have participated in the models

formation in combination with TOPO and 2DAUTO classes. In this process, one more

155


descriptor of TOPO and four more descriptors of 2DAUTO class have joined the

descriptor-lists of respective classes as contributing ones. With respect to activity, these

five descriptors may be carrying marginal information. Interestingly, the participating

empirical and constitutional descriptors also favor more saturated and hydrophobic

structural scaffolds for better activity (Table 4).

All the four participating BCUT class descriptors correspond only to positive

Burden eigenvalue-2. Among these, BEPm2 is correlated with the activity to the order of

0.74. All the four descriptors are intercorrelated (r - 0.60 to 0.99). Many of the

descriptors listed in Table 4 have been associated with negative regression coefficients. It

suggests that activity profile of nitro! acetamido alkenol derivatives (Table 2) and the

identified descriptors have inverse proportionality relationship. If one views the

compounds in terms of a normal distribution curve of their activity with respect to the

molecular descriptors, all these analogues correspond to the post-optimum region of the

said curve. Eq. 8 with BCUT and 2DAUTO class descriptors has emerged as the best

model of dataset 'B' of nitro! acetamido alkenol derivatives (Table 2). This has improved

further on the exclusion of compounds 18 and 19 (Eq. 8a).

-logMIC = - 6.278(2.627)BEPv2 -18.902(3.729)MATS7m + 25.609

n=ll, r=0.884, Q2=0.678, s=0.254, F=14.42 (8)

-.logMIC =- 6.205(1.384)BEPv2 - 19.045(1.978)MATS7m + 25.369

n=9, r=0.973, Q2=O.864, s=0.133, F=52.64 (8a)

For chloro! amino alkenol derivatives (Table 3), the dataset 'B' has resulted in no

better model than those from 2DAUTO class alone. However, due to the synergetic

contributions of a constitutional (CONST) and two Modified Burden eigenvalue (BCUT)

descriptors, two additional models have emerged from this dataset and Eq. 7 is one of

them. The participating CONST descriptor suggests that chloro substitution on the

structural template leads to better activity (Table 6).

156


6.3 QSAR STUDY ON C-3 ALKYL! ARYLALKYL-2,3-DIDEOXY

HEXENOPYRANOSIDES

Inrecent modeling studies, the alignment-free 3D-descriptors from the molecular

geometry are increasingly put to use as an alternative. to the grid-based 3D-QSAR

studies.27-34 They offer insight of the receptor via the conformational space of the probe

molecules and are successfully used in the development of rationales for divergent

biological activities including tubercolosis?8,30,32-36 In view of this, a QSAR of

antitubercular actiVIty of substituted C-3 alkyV arylalkyl-2,3-dideoxy

hexenopyranosidesl9 (Fig. lb, Table 8) is attempted using the alignment-:-free 3D

descriptors from Dragon software.20 The QSAR models have been developed using CP

MLR (combinatorial protocol in multiple linear regression)21 in conjunction with a three

stage descriptor classification protocoe7 and pactialleast square (PLS) method. The study

was attempted with two structure databases representing the X-ray structure 4dr (Figures

2 and 3a) and the conformational analysis derived minimum energy structure of a high

active compound 5c (Fig. 3b). The conformational space of these two structure databases

differ in their C-3 substitution region (Fig. 3). Alignment-free 3D-descriptors from

Dragon software20 were opted as indices of the conformational space. It resulted in 681

and 679 descriptors, respectively, for 4dr and Sc databases. The computational

procedures of embedded structural information divide these indices into eight descriptor

classes (Table 9).20

Among the analogues shown in Table 8, the crystals of 4dr were obtained on

slow evaporation of its solution in acetone-hexane (1:3) at room temperature. Fig. 2 is

ORTEP diagram of 4dr from the X-ray study. It crystallized in monoclinic form with P21

space group. The central hexenopyranose ring is in a distorted half chair conformation.

with unsaturation between C-2 and C-3. In Cahn-Ingold";Prelog system,38 the relative

conformations ofC-I, C-4, C-5 and C-l' are assigned as S, S, R, and R, respectively.

6.3.1 MATERIALS AND METHODS

6.3.1.1 Dataset. Two structure databases of hexenopyranoside derivatives (Table 8),

corresponding to the X-ray structure (4dr) and the minimum energy structure from the

conformational analysis of a high active compound (5c), were created for the

computation of molecular descriptors. The database structures of X-ray structure (4dr)

157


were generated in MOE (molecular operating environment) software39 by making

appropriate changes to the crystal structure template followed by energy minimization in

AMI semi-empirical method40 (rms gradient 0.001) as implemented in the software to

result in the final conformations.

For the second structure database, the minimum energy confomiation of 5c was

identified through its systematic conformational analysis in MOE with default parameter

settings followed by energy minimization (force field: MMFF94)41 and refinement of

final conformation in AMI method. The remaining structures of the database were

generated from this template of 5c by appending appropriate changes to it followed by

structural refinement in AMI method.

Dragon sofiware20 was·used for the computation of alignment free 1D-descriptors

of the structure databases. This resulted in 681 and 679 descriptors, respectively, for the

X-ray (4dr) and the minimum energy conformation (5c) structure databases. The

antitubercular activity (MIC,. in moles per liter) of the compounds· (Table 8) was

considered after its transformation into logarithm of the inverse of inhibitory

concentration (-logMIC). The QSAR models were generated using CP-MLR21 in

conjunction with a three-stage descriptor classification protocoe7 and PLS analysis.42

6.3.1.2 Model development. The CP-MLR/' a filter based variable selection

procedure has been used for the model development. Its procedural aspects and

implementation are discussed in Chapter 3. Throughout this study, for the filters-I, 2,

and 4 the thresholds were assigned as 0.79, 2.0, and 0.3 ~ Q2 ~ l.0, respectively. The

filter-3 was assigned an initial value of 0.71. In order to collect the descriptors with

higher information content, the threshold of filter-3 was successively incremented with

increasing number of descriptors (per equation) by considering the r-bar value of the

preceding optimum model as the new threshold for next generation.

158


Table 8. The observed and predicted antitubercular activity of C-3 alkyl and aryl alkyl

2,3-dideoxy hexenopyranosides (Fig. 1 b)

0yR2

0 Y

OCH(CH3)2

HO

R1

Compd. Rl R2 Y -lorMIC . Obsd.a Eq.l- E9.26 PLS6

4a C6Iit-o-N02 C(CH3)3 =0 3.93 4.05 3.78 4.03 4b C6Iit-m-N02 C(CH3)3 =0 3.93 . 4.06 4.2 3.87 4c C6Iit-p-N02 . C(CH3)3 =0 3.93 4.29 4.29 4.11 4d C6Iit-p-CN C(CH3)3 =0 5.11 4.54 ·4.66 4.83 4e C6Iit-p-F C(CH3)3 =0 4.15 4.27 4.15 4f C(;H4-P~CF3 C(CH3)3 =0 3.95 4.12 4.08 4.04 4g (CH2hCH3 ·C(CH3)3 . =0 4.14 3.98 4.03 4.34 4h (CH2)8CH3 C(CH3)3 =0 4.83 5.02 5.05 5.02 5c C6Iit-p-N02 CH3 =0 5.39 4.94 4.99 5.03 6c C6l4-p-N02 . C6H4-P-N02 =0 4.29 4.08 4.10 4.38 6d C6H4-P-CN C6H.t-p-N02 =0 4.27 4.46 4.78 4.66 6e C6H.t-p-F C6Iit-P-N02 =0 4.26 4.16 4.08 4.16 6f C6H.t-p-CF 3 C6H4-p-N02 =0 4.31 4.13 4.18 4.36 6g (CH2hCH3 C6Iit-p-N02 =0 4.21 4.47 4.17 4.04· 6h (CH2)8CH3 C6Iit-p-N02 =0 4.29 4.38 4.49 4.25 4ar . C6Iit-o-N02 C(CH3)3 -OH 4.23 4.28 3.99 4.00 4br C6H4-m-N02 C(CH3)3 -OH 4.23 3.87 3.97 3.94 4cr C6H.t-p-N02 C(CH3)J -OH 4.44 4.55 4.30 4dr C6Iit-p-CN C(CH3)J -OH 4.60. 4.76 4.85 4er C6H.t-p-F C(CH3)3 -OH 4.20 4.45 4.35 4.28 4fr C6Iit-p-CF3 C(CH3)3 -OH 4.25 4.08 4.31 4.07 4hr (CH2)8CH3 C(CH3)3 -OH 5.44 5.00 4.92 5.19 5cr C6H4~P-N02 CH3 -OH 4.18 4.90 4.46 4.46 6cr C6H.t-P-N02 C6H4-P-N02 -OH 4.29 4.17 4.54 4.56 6fr C6Iit-p-CF3 C6H4-P-N02 -OH 4.31 4.14 4.03 4.26 6hr {CH2)8CH3 C6H4-,e-N02 -OH 4.30 4.36 4.38 4.39

a, Reference 19. b, Activity from corresponding equation.

159


NI

Fig. 2. The ORTEP diagram (30% probability) of 4dr with atomic numbering scheme.

a b

c

Fig. 3. (a) X-ray structure of 4dr; (b) conformational analysis derived minimum energy

structure of 5c; (c) superimposition pose of 4dr (carbons: green) and 5c (carbons:

yellow); in all, oxygens are in red and nitro gens are in blue; hydrogens are suppressed for

clarity.

160


Table 9. The 3D descriptor classes used for the analysis of antitubercular activity of2,3-

dideoxy hexenopyranosides (Table 8)

Descriptor class

(acronymst

Charge descriptors

Aromaticity indices

Definition and scope

Electronic descriptors defined in terms of atomic charges

and describe electronic aspects both of the whole molecule

and of particular regions, such as atoms, bonds, molecular

fragments·

Four aromaticity indices for aromatic compounds

Randk Molecular Profiles RMPs are sequences of molecular descriptors proposed by

(RPM)

Geometrical descriptors

(GEO)

Radial Distribution

Function (RDF)

3D-MoRSE

Weighted Holistic

Invariant Molecular

descriptors (WHIM)

GEometry, Topology, and

Atom-Weights AssemblY

descriptors (GETAWAY)

a, Reference 20.

Randic and derived from the distance distribution moments

of the geometry matrix (interatomic geometric distances)

of a molecule

Conformationally dependent descriptors calculated on the

graph representation of molecules using the geometric

distances between atoms

Molecular descriptors obtained by radial basis functions

centered on different interatomic distances (from o.sA to

Is.sA)

Molecular descriptors calculated by· summmg atom

weights viewed by a different angular scattering function .

Molecular descriptors obtained as statistical indices of the

atoms projected onto the 3 principal components obtained

from weighted covariance matrices . of the atomic

coordinates

Molecular descriptors calculated from the elements of

leverage matrix obtained by the centered atomic

coordinates called Molecular Influence Matrix (MIM)

161


{;\, f:\ f:\ Cat. I

~~0···

• • •

I

aj + .1j

hi + hj

Category I

ai + I)j

IJi + qj c· + I' I. J ~

• V

'> ...

M o o E L S

I

Category III

Ii + Ij

llli+lUj

8 ni ! nj • '- ~

< - / "

0 Ii + IUj

lUi + I1j

IIi + Ij

0 • .. <-

Category II • • •

1.--------...,.------.,. I CategorylV

I Cat. I II Cat. II II Cat. III I 880··· Fig. 4. Schematic representation of categorization of descriptor classes. A, B, C, etc in

circles are descriptor classes and aj, bj, Cj etc, respectively, 'are their constituent

descriptors. Depending on the information content to correlate the activity, the descriptor

classes are put into four categories as primary contributors (category I), collective contributors (category II), secondary contributors (category III) and non-contributors

(category IV). Coefficients of aj , bj , Cj etc are not shown.

6.3.1.3 Descriptor classification protocol. The three-stage descriptor classification

protocoe7 (Fig. 4) is implemented with the 2-desc.riptor combinations (baseline models)

as they are the simplest to understand and explain the activity. In the 1 st stage of

classification protocol, the correlations of the activity with 2-descriptor combinations

from,the individual descriptor classes (DCs) of the dataset were used to sort the DCs into

four categories. They are primary contributors (category I: a DC forms model with its

constituent descriptors), collective contributors (category II: a DC unable to form model

162


with its constituent descriptors, but forms model(s) in combination with a descriptor from

another such DC),. secondary contributors (category III: a DC forms model(s) only in

combination with category I) and non-contributors (category IV: a DC unable to form

model(s) in any manner like category I, II or III). The sorted DCs were collated in the 2nd

stage to identify all the 3-descriptor models across the categories. In the last stage, the

individual descriptors of all 3-descriptor models were pooled to discover the higher

models for the activity.

All the identified models were reassessed for the chance correlations by repeated

randomization of the activity?4,25 Each identified model was subjected to one hundred

simulation runs with scrambled activity and the emerging correlations were counted to

express the percent chance correlation of the model under examination. The proposed

models were validated through two test sets, one from the single linkage hierarchical

clustering of all the descriptor of the compounds and the other from the random selection

procedure, with each containing 8 out of 23 compounds in analysis. The descriptors

identified in the CP-MLR were evaluated through the PLS procedure42 as well. .

6.3.2 RESULTS AND DISCUSSION

The alignment-free 3D-descriptors of X-ray crystal structure (4dr) database of

hexenopyranosides (Table 8) did not lead to any model for the antitubercular activity.

Under identical conditions, three 3D-descriptors classes, namely GEO (Geometrical

descriptors)/o WHIM (Weighted Holistic Invariant Molecular descriptors)43 and

GET A WAY (GEometry, Topology, and Atom-Weights AssemblY descriptorsi,,32 of the

minimum energy conformation of high active compound (Sc) have evolved as collective

contributors (category II) to explain the activity. At the end of a search for four parameter

models, 50 descriptors (Table 10) have emerged from these descriptor classes to model

the activity. The following are selected three- and four-parameter models for the activity

from a pool of equations.

-logMIC = - 3.667(0.558)PJI3 - 5.204(2.11)E3v + 11.832(3.735)Rlu+ + 7.654

n=23 , r=0.835, s=0.251, F=14.61 (1)

163


-logMIC =- 3.571(0.518)PJI3 -14.986(4.940)ISH + 40.044(10.726)HATS2p

-59.740(17.397)R6e+ + 19.29

n=23, r=0.879,· Q2=0.573, s=0.224, F=15.23 (2)

The participating descriptors correspond to these equations are shown in Table II.

In the randomization study, none of the identified models has shown any chance

correlation. These are further validated through two test sets corresponding to descriptors

clustering and random selection procedures with each one containing eight out of twenty

three compounds of Table 8. The test sets predictions are in agreement with their

experimental values (Fig. 5; Table 12).

Among the descriptors, PJI3 (3D Petitjean shape index; GEO class) has the

highest influence on the activity of these analogues. It is a measure of eccentricity in the

molecules. Its regression coefficient suggests that molecular structures with compact

. conformational features are favorable for activity. In Eq. 1, E3v is a directional WHIM

descriptor for axial density (v; weighted by atomic van der Waals volumes) along

principal axes 3. Its regression coefficient suggests in favor of reduced axial density for

enhancement in activity.

The GETAWAY descriptors ISH, Rlu+, R6e+and HAtS2p (Eqs. I and 2) are

embedded with the information from the molecular influence matrix (MIM, H-matrix). In

this class, ISH (standardiZed information content on the leverage equality) encodes

molecular symmetry information. It suggests in favor of symmetry in molecules for better

activity. The descriptors Rlu+ and R6e+, respectively, represent the influence of maximal

autocorrelations of lag I(u; unweighted) arid lag 6 (e; weighted by atomic Sanderson

electronegativities) of R-matrix (MIM as a fraction of distance matrix) on the activity.

The descriptor HA TS2p is autocorrelation of lag 2 (p; weighted by atomic

polarizabilities) from the MIM. the regression coefficients of these descriptors suggest in

favor of increasing number of one and two centered fragments in the chosen molecular

conformation for better· activity. Besides the three- and four-descriptor equations, a

number of higher models also exist among the identified descriptors (Table 10).

164


Table 10. The CP-MLR identified 3D-descriptors of minimum energy conformation (Sc) directed structure database of 2,3-dideoxy hexenopyranosides (Table 8) for modeling

their antitubercular activity

No. Descriptora Reg Coef(fc~ No. Descriptora Reg Coef (fc~ Order Order

Geometrical 1 SPAM 0.281 (0.005)45 25 HATSOv 3.705(0.030)11 2 SPH -1. 760(-0.054)3 26 HATS Iv 2.745(0.018)18 3 PJI3 -1.411 (-0.1 09) 1 27 H6e 0.026(0.005)44 4 DIS Pm -0.006(-0.020)16 28 HATS6e -0.669(-0.016)22 5 DISPe -0.441(-0.042)6 29 Hlp 0.169(0.009)36 6 DISPp -0.046(-0.007)38 30 HATSOp 3.732(0.027)15 7 G(N .. F) 0.001(0.009)37 31 HATSlp 3.018(0.018)19

WHIM 32 HATS2p 0.095(0.001)49 8 Llu -0.010(-0.015)24 33 HATS3p -2.052(-0.014)27 9 L3u 0.120(0.039)7 34 R2u 0.049(0~005)43

10 LIv -0.008(-0.014)26 35 R3u 0.097(0.012)30 11 Glv 16.783(0.056)2 36 R6u -0.023(-0.002)48 12 . G3v 9.568(0.036)9 37 Rlu+ 2.388(0.029) 12 13 E3v -0.622(-0.013)28 38 Rlm+ 0.587(0.017)21 14 E3e -1.148(-0.038)8 39 R2m+ -0 .562(-0.016)23 15 LIp -0.008(-0.013)29 40 R3m+ -5.194(-0.031) 10 16 G2p 1.283(0.006)42 41 R6m+ -1.785(-0.007)41 17 Tp 0.001(0.002)47 42 RTm+ 0.482(0.014)25 18 Gu -16.095(-0.048)4 43 R2v+ 2.188(0.007)39

GETAWAY 44 R3e 0.110(0.010)32 19 ISH -6.043(-0.044)5 45 RTe -0.0002(-

0.0003)50 20 HIC 0.019(0.003)46 46 R2e+ -0.826(-0.017)20 21 H6u 0.051(0.010)31 . 47 R3e+ -3.823(-0.028) 14 22 HTu 0.002(0.009)35 48 R6e+ -8.606(-0.029) 13 23 HATSOu 0.505(0.010)33 49 RIp 0.267(0.009)34 24 HATS6u _ -0.844{-0.019)17 50 R2p+ . 2.1 88{0.007)40

a, Geometrical: SPAM,average span sphere radius; SPH, spherosity; pn3, 3D Petijean shape

index; DISPw, displacement between the geometric centre and the centre of the w property field,

calculated with respect to the molecular principal axes, the last character Ow' refers to the

weighting property which may be either Om' for atomic masses or 'v' for atomic van der Walls

volumes or 'e' for Sanderson electro negativities or 'p' for atomic polarizabilities Or 'u' for

. unweighted; G(N .. F) sum of geometrical distances between Nand F; WIDM: Lnw, Gnw and

Enw, directional indices of size, symmetry and density, respectively; in this, n refers to the

directional component 1,2 or 3 and 'w' refers to the weighting property (for Ow' see DISPw); Tw

and Gw, global indices of size and symmetry weighted by property w; GET A WAY: ISH,

standardized information content on the leverage equality; HIe, mean information content on the

leverage magnitude; Hkw and Rkw, H (Molecular Influence Matrix; MIM) and R (MIMI distance

matrix) autocorrelation descriptors of lag k weighted by property w; lag k refers to the number of

edges in the fragment unit considered in the computation (for Ow' see DISPw); HTu, H total index

165

Functionalized Alkenols & 2,3-Dideoxj Hexenopyranosides as Antitubercular Agents

unweighted; HA TSkw, leverage-weighted autocorrelation of lag k weighted by property w;

Rkw+, R maximal autocorrelation of lag k weighted by property w; see reference 20.

b, MLR like regression coefficient ofthree-component PLS model; (fc) is fraction contribution of

the regression coefficient to the activity; order indicates the order of their significance in the PLS

model; the constant term ofPLS model is 10.328; number of compounds are 23.

Table 11. The descriptors correspond to Eq. 1 and Eq. 2 for 2,3-dideoxy

hexenopyranosides (Table 8)

Compd. 4a 4b 4c 4d 4e 4f 4g 4h 5c 6c 6d 6e 6f 6g 6h 4ar 4br 4cr 4dr 4er 4fr 4hr 5cr 6cr 6fr 6hr

0.859 0.765 0.759 0.717 0.748 0.759 0.891 0.523 0.621 0.81

0.807 0.817 0.853 0.823 0.738 0.794 0.783 0.722

0.75 0.685 0.754 0.454 0.676 0.793 0.855 0.732

0.323 0.347 0.347 0.355 0.346 0.341 0.307 0.311 0.31

0.331 0.34

0.297 0.269 0.274 0.273 0.315 0.347 0.334 0.349 0.331 0.337 0.314 0.316 0.326 0.261 0.274

0.103 0.084

0.1 0.133 0.088 0.085 0.102 0.072 0.109 0.096 0.126 0.089 0.088 0.102 0.071 0.099 0.082 0.099 0.128 0.085 0.082

0.07 0.107 0.095 0.086 0.068

R6e+a

0.024 0.026 0.021 0.021

0.03 0.021 0.017 0.013 0.024 0.023 0.023 0.028 0.024 0.018 0.015 0.023 0.025

0.02 0.018 0.028 0.016 0.013 0.024 0.019 0.017 0.015

1 0.986

1 1

0.985 1

0.982 1

0.969 1

0.988 1

0.968 1

0.989 1

0.986 0.986

1 1 1 1 1

0.989 1

0.989

HATS2pa

0.1 0.098 0.098 0.108 . 0.105 0.093 0.092 0.083 0.102 0.102

0.11 0.11

0.098 0.097 0.088 0.098 0.095 0.095 0.105 0.103 0.091 0.081 0.099

0.1 0.096 0.085

a, pn3, 3D Petijean shape index; E3v,directional index for axial density weighted by atomic van

der Waals volumes along principal axes 3; Rlu+ and R6e+, respectively, represent the influence of

maximal autocorrelations of lag l(u; unweighted) and lag 6 (e; weighted by atomic Sanderson

electronegativities) of R-matrix (MIM as a fraction of distance matrix); ISH, standardized

information content on the leverage equality; HA TS2p, leverage-weighted autocorrelation of lag

2 weighted by atomic polarizabilities.

166

6.0

5.5

-g 5.0 -u "C OJ

0: 4.5

4.0

3.5

6.0

5.5

-g 5.0 -u :0 Q)

D: 4.5

4.0

3.5


'-

I-

t-

~

t-

t-

t-

I I

I Cluster I -

~ .. 0 .. ~'"

'" Mil

B' ",I-.. ,2training= 0.72 2

I test =0.64 , , I

4.0 4.5 5.0 5.5 6.0 Observed

a

I I I I

. I Cluster I .. o .. ..

'" <i~. -

• ",fa o ~training = 0.76

0 2 1 test =0.74

1 , I , 4.0 4.5 5.0 5.5 6.0

Observed C

6.0

5.5

]l5.0 (.)

-a Q)

0: 4.5

4.0

3.5

6.0

]l 5.0 u "C Q)

0: 4.5

4.0

3:5

r-

t-

1-

l-

t-

I I I I

IR.indQIIl I -

~ -'" .&. .. a&-~

; ",1 ~training=0.671-Q

2 I test =0.63

i .1 .1 .1 4.0 4.5 5.0 5.5 6.0

Observed b

I I I I

IRandom I -

"-0 -.. A

• 0 -••

0 t4. .. .~ o • r2training = 0.79 .. 2 =0.69 ( test ~. i

4.0 4.5 5.0 5.5 6.0 Observed

d

Fig. 5. Plots of training ( .. ) and test sets (0) predicted activities versus observed activity corresponding to Eqs. lea, b) and 2 (c, d). Number of compounds in training set is 15 and in test set is 8.

167

FunctionaIized Alkenols & 2,3-Dideoxy Hexenopyranosides as Antitubercular Agents

Table 12. Predicted residual activity of descriptor cluster and random test sets (8

compounds each) of the compounds of Table 8 and the corresponding statistics

Compd. Residualsa

Descri~tor clustcrO Randomc

Eg.l Eg.2 PLS Eg.l Eg.2 PLS 4a -0.17 0.09 -0.45 4b -0.13 -0.32 -0.13 4c -0.31 -0.33 -0.36 4f -0.17 -0.12 -0.34 -0.10 -0.07 -0.31 6c 0.15 0.17 -0.01 6e . 0.10 0.11 0.45 6g -0.26 0.01 0.16 6h -0.04 -0.11 0.01 4br 0.33 0.24 0.25 4hr 0.30 0.39 0.50 0.43 0.46 0.33 5cr -0.66 -0.26 -0.46 6cr 0.12 -0.25 -0.16 6fr 0.14 0.30 0.21 0.11 0.19 0.30 Regression statistics r 0.851 0.874 0.929 0.818 0.891 0,894 Q2 0.429 0.455 0.745 0.395 0.381 0.622 R2Pred ·0.640 0.736 0.516 0.681 0.688 0.532 s 0.234 0.226 0.165 0.267 0.220 0.208 F 9.64 8.11 22.95 7.41 9.7 14.58

a, Difference of observed -logIC5o and predicted -logIC50; each training set has 15 compounds. b, Test set from the descriptors cluster. . C, Test set from random selection.

Training set equations correspond to Table 12 (DC: Descriptor cluster, Rand: Random)

-logMIC = - 3.599(0.721)PJI3 - 4.798(2.548)E3v + 13.929(4.006)R1u+ + 7.306 n=15,r=0.851, Q2=0.429, s=0.233, F=9.64 (DC-I)

-logMIC = - 3.474(0.688)PJI3 - 16.886(5.61l)ISH + 34.861(12.103)HATS2p-42.297(21. 191)R6e+ +21.239

n=15, r=0.874, Q2=0.455, s=0.226, F=8.11 (DC-2)

-logMIC = - 3.329(0.773)PJI3 - 5.692(3.293)E3v + 14.437(4.688)Rlu+ + 7.294 n=15, r=0.818, Q2=0.395, s=0.267, F=7.41 (Rand-i)

-logMIC = - 3.228(0.640)PJI3 -15.589(5.353)ISH +46.718(l2.492)HATS2p-66.787(21. 186)R6e+ + 19.116

n=15, r=0.892, Q2=0.381, s=0.220, F=9.70 (Rand-2)

168


A PLS (partial least squares) analysis has been carried out on these 50 CP-MLR

identified descriptors (Table 10) to facilitate the development of a 'single window'

structure-activity model. In the PLS cross-validation,42 three components are found to be

the optimum for these fifty descriptors (dataset-I) and they explained 76% variance in the

activity (r=0.872, Q2=0.588, s=0.223, F=20.20). A PLS analysis has also been carried out

on the dataset devoid of these 50 identified descriptors (dataset-II; 629 descriptors).

Under identical conditions, the dataset-II could explain only 63% variance in the activity

(r=0.794, Q2=0.345, s=0.278, F=4.55). The dataset of 681 descriptors of the X-ray

structure (4dr) database has resulted in a far inferior PLS model and explained only 57%

variance in the activity (r=0.758, Q2=0.264, s=0.298, F=8.54). These results clearly

substantiate the relevance of the 50 identified descriptors (Table 10) in modeling the

activity of these compounds. The MLR-like PLS coefficients of these 50 descriptors are

given in Table 10. Fig. 6 shows a plot of the fraction contribution of normalized

regression coefficients of these descriptors to the activity (Table 10).

The PLS analysis has also suggested that P J13 as the most determining descriptor

for modeling the activity of the compounds (Table 10; descriptor no. 3 in Fig. 6). The

ot4er significant descriptors in the PLS model are SPH, DISPe (GEO), Gu, Glv, G3v,

L3u, E3e (WHIM), ISH and R3m+ (GETAWAY) (Table 10; descriptor nos. 2,5, 18, 11,

12, 9, 14, 19 and 40 in Fig. 6). Among them, SPH (spherosity) is an anisomertry

descriptor and DISPe is a COMMA229 descriptor. The coefficient SPH suggests in favor

of flat molecules for the activity. DISPe is a measure of displacement between the

geometric and the electronegativity (e) centers in the molecule. Its negative regression

coefficient suggests that an increasing separation between geometric and

electronegativity centers is detrimental to the activity. Both ISH (GETAWAY) and Gu

(global molecular symmetry, unweighted; WHIM) infer about molecular symmetry (Eq.

2; Table 10). They suggest in favor of symmetry in the compounds for the activity. The

directional WHIM descriptors Glv, G3v, L3u and E3e (weighted by v, u'or e) represent

the projections of symmetry (G), dimension (L) and density (E) of the atoms along

respective principal axis 1, 2 or 3. Among these, the coefficients of Glv, G3v and L3u

(Table 10) suggest in favor of larger 1 and 3 axial symmetries and dimension for better

activity. The coefficient of E3e (Table 10) suggests for a reduced axial density, weighed

by electronegativity. The descriptor R3m + (GET AWAY) is R maximal autocorrelation

169


of lag 3 weighted by atomic mass (m). Its coefficient (Table 10) suggests for three

centered structural fragments with minimum atomic masses. Collectively, the analysis

suggests that minimizing ofPll3, SPH, DISPe, Gu, E3e, ISH and R3m+ and maximizing

of G 1 v, G3v and L3u in the compounds would lead to improved activity.

In terms of molecular features, an extended structural frame, reduced symmetry

and separation in geometric and electronegativity centers are detrimental to the activity.

At the same time the compounds with conformational features of enhanced dimension

and symmetries along their principal axis 1 and 3 are favorable for the activity. In

comparison to these ten descriptors, the remaining ones appear in lower order of

significance to influence the activity of the compounds (Table 10; Fig. 6). A brief

. description of all the 50 descriptors is given in Table 10.

0.08

0.06

Effect 0 n activity Fawrable

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 'Jl 39 41 43 45 47 49

Fig. 6. Plot of fraction contribution of MLR-like PLS coefficients (normalized) of the 50

descriptors (Table 10) to the activity. The horizontal axis refers to descriptors' serial

numbers. The ten most significant descriptors are identified by gray shaded lines.

170


6.3.3 EXPLORATION OF MODELS - SYNTHESIS OF NEW

COMPOUNDS

To explore the models several new structures have been identified within the

scaffold of 2,3-dideoxy hex-2-enopyranoside. Structures shown in Table 13 are part of

this explorations predicted by Eqs. 1, 2 and PLS model (Table 13). Of these two

compounds, 5d and 5dr, were selected for synthesis as their limits of predicted activities

are in close range. These compounds have been prepared according to the scheme I and 2

and evaluated for their antitubercular activity.

Table 13. The predicted and observed antitubercular activity of New 2,3-dideoxy hex-2-

enopyranosides (Fig.l b)

6.3.3.1. Chemistry. The synthesis of the compounds 5d and 5dr was initiated with 3,4,6-

tri-O-acetyl-a-D-glucal (1).44 Iodine catalyzed Ferrier rearrangement45,46 of. 1 with

isopropanol furnished 4,6-di-O-acetyl-2,3-dideoxy hex-2-enopyranoside (2). Its

deacetylation yielded dihydroxy derivative (3), which on allylic oxidation47 followed by

acetylation of C-6 hydroxyl group resulted in 4 in good yield (Scheme 1).

The Morita-Baylis-Hillman (MBH) reaction48•5o of 4 with p-cyanobenzaldehyde

in the presence of TiCIJTBAI in DCM at -78 to -30°C yielded highly diastereoenriched

5d (Scheme 2). The reaction is almost completely diastereoselective (>99%) without

using any chiral catalyst. It is due to the formation of an ortho-ester intermediate during

the reaction. Making use of our previously reported procedure, 51 the C-4 keto group of 5d

were stereos electively reduced with NaB~ in the presence of CeCh. 7H20 leading to the

171


formation of their erythro derivative 5dr (Scheme 2). All the compounds were purified

by column chromatography and characterized by elemental analyses, IR, NMR and

MASS spectra.

Scheme 1a

AcO Aco~9 ACO~

1

AcO HO H3CO~CO a. _)~O. b b ~5)~O. c,d 0 0 ACO/~ • HO/~1 ..

- OCH(CH3h 3-20CH(CH3h - OCH(CH3h 2 3 4

a Reagents: (a) (CH3hCHOH, THF, 12, 3h, rt; (b) NaOMe, MeOH, 2h, rt; (c) Mn02, CHCl), 24h,

rt; (d) (CH3CO)20, Pyridine, 6h, O·5°C.

Scheme2a

b ..

5' 5'

4 5d 5dr

a Reagents: (a) TiCI4, TBAI, P·CNC6~CHO, DCM, 5h, ·78°C_ -30°C; (b) NaB~, CeCl) , 7H20,

C2HsOH, 3h, rt.

6.3.3.2 Experimental section. All the reactions were monitored by warming the

CeS04 (1% in 1M H2S04) sprayed precoat6d silica gel TLC plates at 100°C. NMR

spectra were recorded on Bruker Avance DPX 200 FT; Broker Robotics and Broker DRX

300 Spectrometers at 300 MHz eH) and 50 and 75 MHz (BC). ForBC NMR reference

CDCh appeared at 77.11 77.4 ppm, unless otherwise stated. Mass spectra were recorded

on a JEOL SX 102IDA 6000 mass spectrometer using argon/xenon (6 kV, 100 rnA) as

the FAB gas. IR spectra were recorded on Perkin-Elmer 881 and FTIR-8210 PC

Shimadzu Spectrophotometers. Optical rotations were determined on an Autopol III

polarimeter using a 1 dm cell at 28°C in methanol or· chloroform as the solvent;

concentrations mentioned are in gl100 mL. Organic solvents were dried by standard

methods. Aldehydes were purchased from Aldrich and Fluka chemical co.

172


Isopropyl-6-0-acetyl-3-(hydroxy (4-cyanophenyl) methyl)-2,3-dideoxy-a-D-glycero-

hex-:2-enopyranoside - 4- ulose (5d). To a stirred solution of tetrabutylammoniurn

iodide, TBAI, (0.2 mmol) in dry CH2Ch (5 mL) at -78°C was added TiCl4 (1.5 mmol)

dropwise. After stirring for two minutes, a mixture containing enuloside 4 (l mmo!) and

p-cyanobenzaldehyde (2 mmol) in dry CH2Ch (5 mL) was added. The reaction mixture

was slowly warmed to -30°C and stirred till completion of the reaction (5h). A saturated

aqueous solution of sodium bicarbonate was added, followed by filtration through a celite

pad. The organic layer from the filtrate was separated, and the aqueous layer was

extracted with CH2Ch (3x 4ml). The organic layers were then combined, washed with

brine solution, and dried over sodium sulfate. The crude product obtained after

evaporation of the solvent was chroI11atographed to yield the pure compound 5d (66.71%)

(Scheme 2).48-50 Oil. Eluent for column chromatography: ethyl acetate: hexane = 4 : 21,

v/v; Rc = 0.26 (ethyl acetate: hexane = 3 : 7). [a.]o = - 5.330 (c= 0.150, CHCh). IR (neat, ~I 1 . .

cm ) 3452, 2231, 1742. H NMR (300 MHz, CDCh) 0 7.65 (2H, d, J4', 3' or J6', 7' = 8.3

Hz, H-4' & H-6'), 0 7.51 (2H, d, J3',4' or h,6' = 8.2 Hz, H-3' & H-7'), 0 6.63 (lH, dd,h,

I' = 1.1 Hz, h, 1 = 3.7 Hz, H-2), 0 5.6 (lH, S, H-l '),05.41 (lH, d, h 2 = 3.7 Hz, H-l), 0

4.64 (lH, dd, J5, 6a = 4.6 Hz, & J5, 6b = 35 Hz,) 0 4.46 to 0 4.44 (2H, m, H-6a & H-6b), 0

4.01 (IH, sept, J = 6.2 Hz, OCH(CH3h), 0 2.02 (IH, S, OCOCH3), 0 1.26 (3H, d, J= i4 Hz, -OCH(CH3h), 0 l.21 (3H, d, J == 7.2 Hz, -OCH(CH3)2). 13CNMR (CDCh + CC14,

50MHz),0 193.9 (C-4), 0 170.3 (OCOCH3), 0 146.0 (q c), 0140.3 (C-2), 0 138.8 (q c), 0

132.3 (C-4' & C-6'), 0 127.61 (C-3' & C-T), 0 118.4 (q C, -CN), 0 112.0 (q C), 0 92.1

(C-l),o 72.l (C-l '),0 7l.8 (C-5), 0 69.7 (OCH(CH3h), 0 62.4 (C-6), 023.2 (OCOCH3),

o 22.1 & 0 20.7 (OCH(CH3)2). ESI-MS: mlz 359; Found 382 [M + Na+]. HRMS: [M:.

OCOCH3]+, calc. 300.1024, found 300.1016.

Isopropyl-6-0-acetyl-3-[hydroxy (4-cyanophenyl) methyl)-2,3-dideoxy-a-D-erythro-

hex-2-enopyranoside (5dr). To a stirred solution of 5d (Immol) in ethanol (lOmL),

NaBfu (O.5mmole) and CeCh.7H20 (0.5mmole) were added and the reaction mixture

was stirred continuously at room temperature till completion of the reaction (3h). After

completion of the reaction (TLC control), excess NaBfu was neutralized with acetone

and the reaction mixture was concentrated in vacuo to get crude product which on

column chromatography yielded the pure compound 5dr (81.54) (Scheme 2).51 Oil.

Eluent for column chromatography: ethyl acetate: hexane = 1 : 4, v/v; Rc = 0.44, (ethyl

173


acetate: hexane = 1 : 1, v/v). [a]D = 21.33° (c= 0.150, CHCh). IR (neat, cm,l) 3434,

2360, 1732. IH NMR (300 MHz, CDCh) ~ 7.65 (2H, d, J4" 3' or J6" 7' = 8.2 Hz, H-4' & H-

6'), ~ 7.51 (2H, d, ly,4' or h,6'= 8.2 Hz, H-3' and H-7'), ~ 5.58 (lH, d, h 1= 2.8 Hz, H-

2), ~ 5.41 (lH, S, H-l'), ~ 5.12 (lH, d, JI,2 = 2.8 Hz, H-l),~ 4.51 (IH, td, J6a,6b = 12.2

Hz, J6a,5= 10.7 Hz, ha,4= 1.8 Hz, H-6a), ~ 4.15 (lH, d, hb,6a=12.2 Hz, H-6b), ~ 4.17-

~ 3.90 (4H, m, H-5, OCH(CH3h, H-4, -OH), ~ 2.06(S, 3H, OCOCH3), ~ 1.23 (3H, d, J=

6.3 Hz, -OCH(CH3h), ~ 1.16 (3H, d, J = 6.2 Hz, -OCH(CH3h). 13CNMR (CDCh,

50MHz), ~172.3 (OCOCH3), ~ 147.2 (q c), ~142.3 (q C), ~132.6 (C-4 & C-6'), ~ 127.3

(C-3' & C,"7'), ~ 126.5 (C-2)~ ~ 119.1 (q c, CN)~ ~ 111.6 (q C), ~ 93.4 (C-1), ~ 75.8 (C-

1'), ~ 71.2 (C-4), ~ 70.2 (OCH(CH3h), ~ 65.1 (C-5) ~ 64.0 (C-6), ~ 23.9 (OCOCH3), ~

22.3 & ~ 21.2 (OCH(CH3h). ESI-MS: mlz 361, found 384 [M + Na+]. HRMS: [M

OCOCH3t, calc. 302.0969, fou!1d 302.0986.

6.3.3.3 Antitubercular activity. The compounds 5d, 5dr and 4d (a compound from

Table 8) were evaluated for their in vitro antitubercular activity against M. tuberculosis

H37Rv in BACTEC radiometric susceptibility assay. The assay was done using M

tuberculosis H37Rv as test strain. The assay grows M tuberculosis in 7H12 medium

containing 14C labeled palmitic acid as substrate and 14C02 is produced. The amount of

14C02 detected reflects the rate and amount of growth occurring in the vial and is

expressed in term of the "Growth Index" (GI). On addition of an antitubercular drug to

the medium suppression of growth of the test organism M tuberculosis can be detected

by either a decline or very small increase of the daily GI output as compared to the

control. Streptomycin (6Ilg/ml) and Rifampin (2llg/ml) were used as positive control.

Compounds 5d, 5dr (Table 13)and 4d (Table 8) have shown 6.25, 50.00 and 6.25 Ilg/ml

as MIC in this test protocol. In agar dilution method, compound 4d has shown an MICof

3.12 Ilg/ml. A linear comparison of compound 4d's activity inBACTEC and agar

dilution methods· led to suggest that compounds 5d and 5dr show respectively 3.12 and

25.00 Ilg/ml as MIC in agar dilution method (Table 13).

The observed antitubercular activity of 5d (3.12 Ilg/ml) is within the predicted

activity range (2.11 - 5.07 IlglmL; Table 13) as suggested by models (Eqs. 1,2 and PLS

model). In 5d the RI and R2 groups are C6H4-p-CN and CH3 respectively and the Y

• The in-house antitubercular activity assay system has been revised to BACTEC radiometric susceptibility assay method from agar dilution procedure used for compounds of Table 8.

174


functional group is C=O. The 5dr is prepared by reduction of carbonyl function group at

Y position of the 5d. The observed activity of 5dr (25.00 J.1g/ml) is out side the predicted

activity range (2.63 - 9.04 J.1g/mL; Table 13). Interestingly, the MIC of all reduced

analogues of this series (Table 8) is 25 J.1g/mL except 4hr (1.56 J.1g/mL, R2 functional

group is (CH2)sCH3). Also the observed MICs of other analogues of this series with R2 =

CH3 i.e. 5c and 5cr are 1.56 and 25 J.1g/mL respectively (Table 8). This shows that the

developed QSARmodelsare robust enough to predict the activity of2,3-dideoxy hex-2-

enopyranosides.

6.4 SUMMARY AND CONCLUSIONS

This study has provided a rational approach for the development of novel

antitubercular agents based on functionalized alkenol and 2,3-dideoxy hex-2-

enopyranoside carbohydrate synthons. The study on functionalized alkenols shows that

the antitubercular activity of both nitrol acetamido alkenol derivatives (Table 2) and

chlorol amino alkenol derivatives (Table 3) have been found to be correlated to simple

(TOPO) to complex (2DAUTO) topological descriptors of these compounds. The

descriptors from other classes, that is, empirical, constitutional and remaining topological

have made secondary contribution in association with TOPO and I or2DAUTO classes.

Exclusion of compounds 18 and 19 from the dataset of nitrol acetamido alkenol

derivatives (Table 2) and compound 23 from that of chlorol amino alkenol derivatives

(Table 3) has improved the significance of all identified models. However, these

exclusions did not skew the order and magnitUde of the regression coefficients of the

descriptors. This . lends credence to· the models and descriptors identified therein. The

antitubercular activities predicted. by the selected models of this study are in agreement

with the observed ones (Tables 2 and 3). In case of nitro/ acetamido alkenol derivatives

(Table 2), the correlations obtained with the TOPO descriptors suggest that less branched

and saturated structural templates would be better for the activity. For both the series of

compounds, in 2DAUTO descriptors, the activity has been correlated to the

autocorrelation vectors with mass, volume and! or polarizability as weighting component..

While most of the models of nitro/ acetamldo alkenol derivatives (Table 2) involved

descriptors of lag one, six and seven, for chlorol amino alkenol derivatives (Table 3) this

happened to be from lag four. Four descriptors (ATS3p, ATS7p, ATS3v and ATS7v) out

175

Functionalized AlkenoIs & 2,3-Dideoxy Hexenopyranosides as Antitubercular Agents

of the six of 2DAUTO class of chloro/ amino alkenol derivatives (Tables 3 and 6) also

appear in the descriptor list of nitro/ acetamido alkenol derivatives (Tables 2 and 4).

In these tWo series of compounds, however, the regression coefficients of these

descriptors have opposite arithmetic signs with respect to one another. Also, unlike the

nitro/ acetamido alkenol derivatives (Table 2), for chloro/ amino alkenol derivatives

(Table 3) most of the contributing descriptors have been associated with positive

regression coefficients. Outwardly these two series of compounds appear very similar.

But in terms of activity they belong to different segments of descriptor-activity profiles.

This difference in the activity of these two series of compounds may be mainly due to the

spacing difference between the C 1 (also C6) substituents and rest of the functional groups

in them (Fig. la). This understanding will be helpful in modulating the spacing between

the functional groups.

The study on substituted 2,3-dideoxy hexenopyranosides (Fig. 1 b, Table 8) have

been carried out using the alignment~free 3D-descriptors from Dragon software. The

study was attempted with two structure databases representing the X-ray structure 4dr

(Figures 2 and 3a) and the conformational analysis derived minimum energy structure of

a high active compound Sc (Fig. 3b). The conformational space of the minimum energy

structure from confoI'mational ~alysis (Sc) and the X-ray structure (4dr) differ in their

C-3 substitution region (Fig. 3). The correlation of the activity of 2,3-dideoxy

hexenopyranosides with the 3D.:.descriptors emanating from the minimum energy

conformation of Sc suggests its relevance to the antitubercular activity. The geometrical

(GEO) descriptors suggest that the molecular structures with compact and! or flat

conformational features are favorable forthe activity_ The involvement of the directional

WHIM descriptors in the correlations suggests the complex nature of interaction between

the compounds and their target. The autocorrelation descriptors from GET A WAY

suggest in favor of 3-centered structural fragments. Collectively, the identified

descriptors suggest in favor of compounds embedded with some degree of symmetry,

compact conformational features and closely placed geometric and electronegativity

centers for enhanced antitubercular activity. To explore the models for 2,3-dideoxy

hexenopyranoside scaffold, two compounds, Sd and Sdr were selected for synthesis and

evaluated for their antitubercular activity. The observed activity of these compounds is in

agreement with the model predictions.

176


The descriptors identified in the study contribute to the understanding of the

designing of new/ novel compounds in terms of optimum molecular features and

physicochemical properties to be associated with them. Among the high active

compounds, 4d has shown no toxicity. It is identified for further development.

PLS factor scores, loadings, weights and sensitivity of independent and dependent

descriptors of the PLS model is given in Annexure 1.

177


6.5 REFERENCES 1. Somoskovi, A.; Parsons, L. M.; Salfinger, M. Respir. Res. 2001,2, 164-168.

2. Khasnobis, S.; Escuyer, V. E.; Chatterjee, D. Expert Opin. Ther. Targets. 2002, 6,

21-40.

3. Jones, P. B.; Parrish, N. M.; Houston, T. A.; Stapon, A; Bansal, N. P.; Dick, J. D.;

Townsend, C.A J. Med Chern. 2000,43,3304-3314.

4. Pasqualoto, K. F. M.; Ferreira, E. I. Curro Drug Targets. 2001,2,427-437.

5. Teodori, E.; Dei, S.; Scapecchi, S.; Gualtieri, F. II Farrnaco. 2002,57,385-415.

6. Frieden, T. R.; Sterling, T. R.; Munsiff, S. S.; Watt, C. 1.; Dye, C. Lancet. 2003, 362,

887-899.

7. Smith, C. V.; Sharma, V.; Sacchettini, J. C. Tuberculosis. 2004, 84, 45-55.

8. Bayles,K. W. Trends Microbiol. 2000,8,274-278.

9. Katz, A. H.; Caufield, C. E. Curro Pharrn. Design. 2003, 9, 857-866.

10. McNeil, M.; Wallner, S.J.; Hunter, S.W.; Brennan, PJ. Carbohydr. Res. 1987, 166,

299-308.

11. Daffe,M.; Brennan, P. 1.; Mcneil, M. J Bioi. Chern. 1990, 265, 6734-6743.

12. (a) Pathak, A. K.; Pathak, V.; Seitz,L.; Maddry, 1. A.; Gurcha,S. S.; Besra, G. S.;

Suling, W. J.; Reynolds, R. C. Bioorg. Med Chern. 2001,9,3129-3143. (b) Pathak,

A. K.; Pathak, V.; Maddry, J.A.; Suling, W. J.; Gurcha, S. S,; Besra, G. S. Reynolds,

R. C. Bioorg. Med Chern. 2001,9,3145-3151.

13. Wen, x.; Crick, D. C.; Brennan, P. J.; Hultin, P. G. Bioorg. Med Chern. 2003, 11,

3579-3587.

14. Centrone, C. A.; Lowary, T. L. Bioorg. Med Chern. 2004, 12, 5495-5503.

15. Cren, S.; Gurcha S. S.; Blake AJ.; Besra, G. S.; Thomas N. R. Org. Biornol. Chern.

2004,2,2418-2420.

16. (a) Couladouros, E. A.; Strongilos, A. T. Angew. Chern., Int. Ed Engl. 2002, 41,

3677-3680. (b) Georgiadis, M. P.; Couladouros, E.A.; Delitheos, A K. J Pharrn.

Sci. 1992,81, 1126-1131.

17. Pathak, R.; Shaw, A. K.; Bhaduri, A. P.; Chandrasekhar, K. V. G.; Srivastava, A.;

Srivastava, K. K.; Chaturvedi, V.; Srivastava, R.; Srivastava, B. S.; Arora, S.; Sinha,

S. Bioorg. Med Chern. 2002, 10, 1695-1702.

178


18. Pathak, R; Pant, C. S.; Shaw, A. K.; Bhaduri, A. P.; Gaikwad, A. N.; Sinha, S.;

Srivastava, A.; Shrivastava, K. K.; Chaturvedi, V.; Srivastava, R; Srivastava, B. S.

Bioorg. Med Chem. 2002,10,3187-3196.

19. 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. 2007,50,2942-2950.

20. Dragon software (version 1.11-2001 and 3.0-2003) by Todeschini, R; Consonni, V.

Milano, Italy and references cited therein. http://www.talete.mi.it.

21. Prabhakar, Y. S. QSAR Comb. Sci. 2003,22,583-595.

22. ChemDraw Ultra 6.0 and Chem3D Ultra, Cambridge Soft Corporation, Cambridge,

USA.

23. HyperChem Release 3, 1993, Autodesk, California, USA.

24. So, S. S.; Karplus, M. J. Med Chem. 1997, 40,4347-4359.

25. Prabhakar, Y: S.; Solomon, V. R; Rawal, R K.; Gupta, M. K.; Katti, S. B. QSAR

Comb. Sci. 2004, 23, 234-244.

26. Broto, P.; Moreau, G.; Vandycke, C. Eur. J. Med Chem. 1984, 19,66-70.

27. Todeschini, R, Lasagni, M. and Marengo, E . .1 Chemometrics. 1994, 8, 263-272.

28. Silverman, B. D.; Platt, D. E. J. Med Chem. 1996,39,2129-2140.

29. Silverman, B. D. J. Chem. In! Comput. Sci. 2000,40, 1470-1476.

30. Pastor, M.; Cruciani, G.; McLay, I.; Pickett, S.; Clementi, S . .1 Med Chem. 2000, 43,

3233-3243.

31. Consonni, V.; Todeschini, R; Pavan, M. J. Chem. In! Com put. Sci. 2002, 42, 682-

692.

32. Consonni, V.; Todeschini, R; Pavan, M.; Gramatica, P . .1 Chem. In! Comput. Sci.

2002,42,693-705.

33. Bagchi, M.C.; Maiti, B.C.; Mills, D.; Basak, S. C. J. Mol. Model. 2004,10,102-111.

34. Bagchi, M.C.; Mills~ D.; Basak, S. C. .1 Mol. Model. 2007, 13, 111-120.

35. Cratteri, P.; Romanelli, M. N~; Cruciani, G.; Bonaccini, C.; Melani, F . .1 Comput.

Aided Mol. Des. 2004, 18, 361-374.

36. Gonzalez, M.P.; Suarez, P.L; Fall, Y.; Gomez, G. Bioorg. Med Chem. Lett. 2005, 15,

5165-5169.

37. Gupta, M. K.; Prabhakar, Y. S. J. Chem. In! Model. 2006,46,93-102.

179


38. Cahn, R. S., Ingold, c.; Prelog, V. Angew. Chem. Int. Eng!. 1966,5,385-415.

39. MOE: The Molecular Operating Environment from Chemical Computing Group Inc.,

1255 University Street, Suite 1600, Montreal, Quebec, Canada H3B 3X3.

http://www.chemcomp.com.

40. Dewar, M. J. S.; Zoebisch, E. G.; Healy, E .. F.; Stewart J. J. P. JAm. Chem. Soc.

1985,10~3902-3909.

41. Halgren, T. A. J Comput. Chem. 1996,17,490--519.

42. Wold, S. Technometrics. 1978, 20, 397-405.

43. Todeschini, R, Lasagni, M.; Marengo, E. J. Chemom.1994, 8, 263-272;

44. Roth, W.; Pigman, W. In Methods in Carbohydrate Chemistry; Whistl er, R L.,

Wolfrom, M. L., Eds.; Academic Press Inc.: New York, 1963; Vol. 11, pp 405-407.

45. Ferrier, R J. Substitution-with-allylic-rearrangement reactions of glycal derivatives.

in Topics in current chemistry; Springer-Verlag: Berlin I Heidelberg, 2001; VoL

215, pp 153-175 and references cited therein.

46. Prevost, N.; Rouessac, F. Synth. Commun. 27,1997,2325-2335.

47. Fraser-Reid, B.; McLean, A.; Usherwood, E. W.; Yunker, M. Pyranosiduloses. II.

Can. J. Chemistry1970, 48, 2877-2884.

48. (a) Morita, K-i.; Suzuki, Z.; Hirose, H. Bull. Chem. Soc. Jpn. 1968, 41,2815. (b)

Baylis, A. B.; Hillman, M. E. D. German Patent 2155113, 1972; Chem. Abstr. 1972,

77, 34174q

49. (a) Basavaiah, D.; Rao, A. J.; Satyanarayana, T. Chem. Rev. 2003, 103, 811-891. (b)

Langer, P. Angew. Chem. Int. Ed. 2000, 39, 3049-3052. (c) Ciganek, E. The

Catalyzed a-hydroxyalkylation and a-aminoalkylation of activated olefins (The

Morita-Baylis-Hillman Reaction) In Organic Reactions: Paquette, L.A., Ed.; Wiley:

New York; 1997, Vol. 51, pp 201-350 (d) Basavaiah, D.; Rao, P. D.; Hyma, R S.

Tetrahedron. 1996, 52, 8001-8062. (e) Drewes, S. E.; Roos, G. H. P. Tetrahedron.

1988,44,4653-4670.

50. Pathak, R; Shaw, A. K.; Bhaduri, A. P. Tetrahedron. 2002,58,3535-3541.

51. (a) Sagar, R.; Pant, C. S.; Pathak, R.; Shaw, A. K. Tetrahedron. 2004, 60, 11399-

11406. (b) In 50a the MBH-adduct derived reduced products were reported as in

threo conformation. Now they are revised as in erytho conformation based on the X

ray analysis of compound 4dr. (c) Sagar, R., Saquib, M., Shaw, A. K., Gaikwad, A.

180


N., Sinha,. S. K., Srivastava, A., Chaturvedi, V., Manju, Y. K., Srivastava, R.,

Srivastava, B. S. C-3 Alkyl or arylalkyl substituted 2,3-dideoxy glucopyranosides

and a process for preparation thereof. Indian Patent, 2006, No. 0533DEL2006, Ref

No. 0210NF2005/IN, 2006.

181

Documents

CFiapter - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/14408/13/13_chapter 6.p… · compounds were first of their class to show this kind of activity . . Here, establishing