ª 2010 John Wiley & Sons A/S

Chem Biol Drug Des 2011; 77: 57–62 doi: 10.1111/j.1747-0285.2010.01045.x

Research Article

Quantitative Structure Activity Relationship

Study of 2,4,6-Trisubstituted-s-triazine
Derivatives as Antimalarial Inhibitors of
Plasmodium Falciparum Dihydrofolate Reductase
Himanshu Ojha, Pragya Gahlot, Anjani K. target DHFR of the parasite, leading to decreased levels of fully
Tiwari, Mallika Pathak, Rita Kakkar* reduced tetrahydrofolate, a necessary cofactor in important one-car-
bon transfer reactions in the pyrimidine, purine, and amino acid bio-
Department of Chemistry, University of Delhi, Delhi 110007, India synthetic pathways (2). The lower levels of tetrahydrofolate result in
*Corresponding author: Rita Kakkar, rkakkar@chemistry.du.ac.in decreased conversion of glycine to serine, reduced methionine synthe-
sis, and lower thymidylate levels, with a subsequent arrest of DNA
replication. DHFR has received considerable attention, as it is the tar-
This study presents a quantitative structure activ- get of pyrimethamine and other antifolates used for the prophylaxis
ity relationships (QSAR) study on a pool of 19 bio- and treatment of P. falciparum infection (3,4). Thus, it becomes imper-
active s-triazine compounds. Molecular descrip- ative to design new molecules possessing therapeutic value and ⁄ or
tors, kappa {1j}, chi {3v}, x component of the dipole improving the existing drugs to fight the resistance of the parasite. In
moment (lx), Coulson charge (qN) on the nitrogen this direction, various classes of nitrogenous compounds, such as
atom sandwiched between the two substituted
substituted triazines (5), pyrimidines (6), and quinolines (7), have been
carbons of the triazine ring, and total energy (ET)
obtained from AM1 calculations provide valuable
synthesized and screened for antimalarial activity. S-triazine deriva-
information and have a significant role in the tives have also received considerable attention because of their
assessment of dihydrofolate reductase (DHFR) potent biological activity, for example, as antiprotozoals (8), antican-
inhibitory activity of the compounds. By using the cer drugs (9), estrogen receptor modulators (10), cyclin-dependent
Genetic Function Approach (GFA) technique, five kinase inhibitors (11), antivirals (12), and antimalarials (13).
QSAR models have been drawn up with the help
of these calculated descriptors and DHFR inhibi- Quantitative structure activity relationships (QSAR) studies are tools
tory activity data of the molecules. Among the for predicting end-points of interest in organic molecules acting as
obtained QSAR models presented in the study, sta-
drugs (14). Many physiological activities of molecules can be related
tistically the most significant one is a four-param-
to their composition and structures. Molecular descriptors, which
eter linear equation with the Lack-of-Fit value
0.5624, squared correlation coefficient R 2 value of are the numerical representations of the molecular structures, are
0.7697, and the squared cross-validated correla- used in QSAR analysis (15).
2
tion coefficient RCV value of 0.6469. The results
are discussed in light of the main factors that We have attempted to build QSAR models using the Genetic Func-
influence the DHFR inhibitory activity. tion Approximation (GFA) method to explore the correlations
between the calculated molecular descriptors of some 2,4,6-trisub-
Key words: antifolates, antimalarials, GFA approach, QSAR, stituted-1,3,5-triazine compounds and their experimental minimum
s-triazines inhibitory concentrations (5) against the P. falciparum DHFR. Based
on the correlations, we have identified the key properties. These
Received 11 September 2009, revised 23 July 2010 and accepted for
publication 19 September 2010 properties can subsequently be utilized for the manipulation of
structural features in the s-triazine moiety in a bid to achieve tri-
azine derivatives with higher potency.
Malaria is a disease that affects a large number of people and
is responsible for a large number of deaths every year. Plasmodium
falciparum is the causative agent of the most lethal form of malaria. Experimental Section
Inhibitors of dihydrofolate reductase (DHFR) have been mainstays in
the treatment of falciparum malaria (1). The enzyme catalyzes the Quantitative structure- activity relationship
nicotinamide adenine dinucleotide phosphate (NADPH)-dependent analysis
reduction of dihydrofolate to tetrahydrofolate. This step is vital for the We have performed QSAR studies on a series of 19 triazine deriva-
proliferation of the parasite. Thus, compounds of this class primarily tives. For this purpose, various molecular properties were calculated

57
Ojha et al.

and these were correlated with the biological activity to obtain vari- Table 1: Antimalarial activities of triazine derivatives (a–s)
ous QSAR models. observed as MIC values
R
A model may have one or more outputs and these may be descrip- N
H
tors, physical properties, or dependent variables. The descriptors, as O2N N N
calculated in this study, fall into two categories: quantum mechani- N
cal and fast descriptors. The quantum mechanical descriptors were R
calculated using the AM1 semiempirical method. The descriptors,
the component of the dipole moment component in the x direction Compound R MIC (lM)
(lx), Coulson charge on nitrogen (qN), and total energy, ET (eV) are (a) 2.418
N
quantum mechanical descriptors.
N
Me
Fast descriptors are a set of efficient algorithms that can be used
to calculate a variety of 2D molecular properties. These are further (b) 3.536
N
classified into Topological descriptors, Structural descriptors, Ther-
modynamic descriptors, Information-content descriptors, and Electro- N
topological descriptors. 1j and 3v fall into the topological class of
fast descriptors, which are 2D descriptors based on graph theory
concepts. These indices have been widely used in quantitative
structure property and quantitative structure activity relationship
(QSPR and QSAR) studies. They help to differentiate molecules (c) N 18.600
according to their size, degree of branching, flexibility, and overall
N
shape. The 1j and 3v indices fall into the Kier and Hall class of
molecular connectivity indices (16). 3v emphasizes different aspects
of atom connectivity within a molecule. Basically, it explains the
influence of clustering in the compound on its activity. 1j is a Kier (d) 2.157
shape index that intends to understand the different aspects of H N
molecular shape. N N

The geometries of all molecules involved in this study were fully (e) H 4.224
N
optimized using the semiempirical AM1 method (17) using the N
MOPAC 7.1 package (openmopac.net). The descriptor set was O
chosen to capture important geometric, electronic, and structural
features. The MS-Modeling 4.4 software (Accelrys, Inc. San Diego, (f) O 3.982
CA, USA) was used for obtaining the parameters and for the H
N N
statistical analysis.
(g) H 2.430
N
Statistical analysis
The GFA analysis was initiated by building a population of 100 ran-
domly constructed equations. This initial population was then H
(h) 4.129
evolved up to a given number of generations. The number of popu- N
lations (i.e. 4000 in this study) was chosen such that the values
of the squared correlation coefficient (R 2) and the squared
2
cross-validated correlation coefficient (RCV ) converged. The number
(i) H 5.564
of descriptors was limited to 20–25% of the size of the data set, N
i.e., four. Finally, we obtained the best five QSAR equations based (CH2)3
on Friedman Lack-of-Fit (LOF) scores (18) for each GFA model. CH3

(j) H 21.202
N
Results and Discussion (CH2)7
CH3
The minimum inhibitory concentrations obtained from the literature
(5) are expressed in lg ⁄ mL. To have more meaningful biological (k) C2H5 27.822
activity data, these were first transformed to molar concentrations.
N
The chemical structures and their minimum inhibitory concentrations C2H5
(MIC, in lM) of 19 triazine derivatives, taken from the literature (5),
are given in Table 1. The data in Table 1 reveal a large variation in

58 Chem Biol Drug Des 2011; 77: 57–62

 QSAR Study of 2,4,6-trisubstituted-s-triazines

Table 1: (Continued ) and performed a quantitative structure activity analysis of the data.
Eighty such parameters were calculated, but, because the number
Compound R MIC (lM)
of compounds is limited to only 19, we applied a GFA to reveal the
(l) H 139.109 important parameters affecting the antimalarial activity. Moreover,
N the data on MIC do not follow a normal distribution. Accordingly,
C(CH3)3
the data were transformed to their logarithms.
(m) 23.393
H The GFA method, implemented in regression analysis, was
N employed for selecting the 'best' regression models, and these are
given in Table 2. The equations are arranged according to decreas-
(n) H 22.968
N F ing LOF values. The 'goodness of fit' of the models was tested on
the basis of the squared correlation coefficient (R 2). For testing the
2
predictive performance of the models, RCV 'leave one out' (LOO), the
squared cross-validated coefficient method was used. The LOO
(o) 20.768 approach consists in developing a number of models with one sam-
N ple omitted at a time. After developing each model, the omitted
N data are predicted and the differences between the experimental
and predicted activity values are calculated. All equations show sig-
(p) N 25.813 nificant regression (Table 2), as evident from their significance of
O regression F-values.

(q) N 130.395 Among the models shown in Table 2, the best scoring model is
Model 1 with LOF = 0.5564, R2 = 0.7722. This is a four-parameter
regression equation. Interestingly, Model 1 does not show the
2
(r) 140.682
best predictive power with RCV = 0.5460. According to Model 1,
log(1 ⁄ MIC) is inversely related to 3v, total energy, E T, x component
N of the dipole moment, lx, and the Coulson charge on the nitrogen
atom (N3), qN (Figure 1). In other words, the minimum inhibitory
concentration increases with increase in 3v, E T, lx, and qN.
(s) H 149.111
N However, as far as predictive power goes, Model 2 is superior, with
OH R 2 only marginally smaller (0.7697), but RCV 2
clearly superior. In
1
MIC, minimum inhibitory concentration for the development of ring stage
this equation, j replaces the total energy in the equation and
parasite into the schizont stage during 40-h incubation. log(1 ⁄ MIC) increases directly with this parameter. Figure 2
illustrates the agreement between the predicted values from this
the activities. While 8 of the 19 compounds show potent activity equation and the experimental values for Model 2. Similarly, the
(MIC < 6 lM), seven are less potent (MIC 18–28 lM), and the rest other four model equations also contain four descriptors each,
are practically inactive (MIC > 100 lM). It is observed from Table 1 which are similar to the descriptors mentioned above, in terms of
that short aliphatic chains (compare (i) and (j)) lead to higher activ- the structural field they represent, and are different combinations
ity. However, reduction of chain length to ethanolamine, as in (s), of the parameters in Models 1 and 2 or their variations, except the
decreases the activity to 149 lM. Branching of the alkyl chain also occurrence of the Kier and Hall third order subgraph counts-cluster
does not favor activity, as the data for (k) and (l) reveal. Compari- (3SC) and octupole moment in Model 3, and the Mulliken charge on
son of (d) and (e) reveals the better efficacy of the imidazole moi- nitrogen (qN') in Model 5.
ety over the morpholine moiety with respect to antimalarial activity.
As far as the contribution of different properties to the respective
To introduce a quantitative aspect to these conclusions, we calcu- equations is concerned, the Coulson charge on the nitrogen contrib-
lated various structure-related parameters for each of the molecules utes heavily in comparison with other properties. The next impor-

Table 2: Obtained quantitative structure activity relationships models for the molecules studied

Statistical characteristics

S. No. Equations Lack-of-Fit R2 2

RCV F-valuea

1 )0.95913v ) 0.0006338ET ) 0.09597lx ) 42.19qN ) 18.00 0.5564 0.7722 0.5460 11.86

2 0.12111j ) 0.94983v ) 0.1028lx ) 34.39qN ) 14.83 0.5624 0.7697 0.6469 11.70
3 )0.25033SC ) 0.0009776ET ) 0.001776Oyyy ) 40.66qN ) 18.10 0.5674 0.7677 0.6160 11.57
4 0.12101j ) 0.94823v ) 0.1029lx ) 34.07qN ) 14.67 0.5712 0.7661 0.6399 11.47
5 0.12471j ) 0.95383v ) 0.1036lx ) 33.46q'N ) 15.88 0.5674 0.7595 0.6317 11.05
a
Critical SOR F-value (95%) = 3.16.

Chem Biol Drug Des 2011; 77: 57–62 59

Ojha et al.

Table 3: Calculated molecular descriptors and predicted and

O experimental inhibitory activity data of compounds studied

N log(1 ⁄ MIC)
O
1 3
Compound j v lx qN Expt. Predicted

(a) 22.74 2.152 )7.617 )0.3478 )0.3835 )0.4056

(b) 31.54 2.391 )4.606 )0.3294 )0.5485 )0.5451
(c) 29.62 2.242 5.137 )0.3396 )1.2695 )1.2289
N1 (d) 26.54 1.725 2.495 )0.3436 )0.3340 )0.7576
(e) 26.54 1.725 )0.390 )0.3385 )0.6257 )0.6371
C1 (f) 28.49 1.725 )1.341 )0.3302 )0.6007 )0.5886
(g) 22.82 1.725 )5.011 )0.3484 )0.3856 )0.2673
N2 N2′
(h) 24.75 1.725 1.822 )0.3357 )0.6560 )1.1786
(i) 21.93 1.316 8.994 )0.3557 )0.7454 )1.2865
C2 C2′ (j) 29.87 1.316 8.704 )0.3390 )1.3264 )0.8692
R N3 R (k) 21.93 1.575 9.308 )0.3626 )1.4444 )1.3164
(l) 21.93 4.438 )2.426 )0.3666 )2.1434 )2.1112
Figure 1: 2,4,6-trisubstituted-1,3,5-triazine derivatives. (m) 24.75 1.725 5.133 )0.3556 )1.3691 )0.9395
(n) 24.62 2.302 7.889 )0.3499 )1.3611 )1.5151
(o) 25.82 2.242 5.630 )0.3497 )1.3174 )1.4215
tant parameter is 3v. On perusal of Table 3, it is observed that the (p) 20.82 1.575 5.147 )0.3419 )1.4118 )1.7257
1
j contribution, wherever present, remains almost the same in (q) 20.89 1.575 2.951 )0.3419 )2.1153 )1.5012
(r) 18.98 1.575 5.012 )0.3402 )2.1483 )2.0053
every model. 3v, which denotes clustering in the compounds, con-
(s) 19.88 1.316 6.198 )0.3319 )2.1735 )2.0590
tributes significantly and consistently, with coefficient magnitude
changing marginally from 0.9482 to 0.9591. Thus, 3v remains an
important property to determine the activity, because 3v and mini-
mum inhibitory concentrations are directly correlated to each other Thus, according to our models, the inhibition activity of these
and clustering contributes consistently and significantly to the activ- compounds is affected by both topological as well as quantum
ity. The Coulson charge on N3 (Figure 1) shows greater variation, mechanical electrostatic properties. Topological properties like
the value of the contribution increasing from 33 in Model 5 to 42 molecular connectivity in terms of 3v or clustering in the compound,
in Model 1. Such large variation indicates that the Coulson charge and quantum mechanical electrostatic properties like Coulson
on nitrogen remains the most powerful deciding factor in estimating charge on the N3 atom of the triazine aromatic ring, are the impor-
the biological activity. The next significant property that appears is tant factors. Thus, the charge density on the nitrogen atom in the
the x component of the dipole moment. The magnitude of the con- lower part of the molecule (Figure 1) is important for the inhibition
tribution of the x component of the dipole moment remains close to activity. The two rings are in the same plane, and so conjugation
0.1. The point to be noted is that these properties contribute con- with the triazine ring will directly affect the charge density on the
sistently with a moderate magnitude, almost equivalent to 1j, to other aromatic ring, and thus the inhibition activity. This is affected
the activity, while other properties like the total energy and by changing the substituents on the carbon atoms. Molecular shape
octupole moment contribute insignificantly to the activity. aspect in term of 1j and polarity contribute mildly to the biological

D : Equation 2: predicted values

–0.2
D : Equation 2: predicted values y = 0.769742 * x – 0.270971
–0.4

–0.6

–0.8

–1.0

–1.2

–1.4

–1.6

–1.8

–2.0 Figure 2: Plot of observed

log(1 ⁄ MIC) values along with the
–2.2
–2.2 –2.1 –2.0 –1.9 –1.8 –1.7 –1.6 –1.5 –1.4 –1.3 –1.2 –1.1 –1.0 –0.9 –0.8 –0.7 –0.6 –0.5 –0.4 –0.3 predicted values according to Eq.
A : Actual values for E : log(1.MIC) (1).

60 Chem Biol Drug Des 2011; 77: 57–62

 QSAR Study of 2,4,6-trisubstituted-s-triazines

activity of the 2,4,6-trisubstituted-1,3,5-triazine compounds as antif- activity. In short, the minimum inhibitory concentration of derivatives
olate agents. It is interesting to note that log P, the indicator of can be improved (i.e. decreased) if such structural features are
lipophilicity, does not find mention in any of these equations. included that will decrease 3v, the Coulson charge of N3, x compo-
nent of the dipole moment, while at the same time increase the
Using Model 2, the predicted inhibition constants of the compounds value of the 1j descriptor. The implication of this study is that the
are presented in Table 3. Out of all the molecular (topological, ther- QSAR method successfully predicts the molecular properties that
modynamical and structural) descriptors derived from our calcula- primarily influence the antimalarial activity of these s-triazine deriv-
tions based on the GFA model of regression analysis, the important atives. With the help of Model 2, the predicted antimalarial activity
influential molecular descriptors for the 19 triazine derivatives com- for newly designed s-triazine derivatives can be calculated. This will
pounds and their experimental and predicted MIC values are pre- eventually help in saving time and valuable resources by synthesiz-
sented in Table 3. These are the descriptors used to select the ing only the most potent s-triazine derivatives and, thus, expedite
dominant parameters affecting the inhibitory activity of the com- the drug designing process. In future, this study should help in
pounds. directing the synthesis of most potent biological s-triazine deriva-
tives and improving the drug designing process and also will be
In the QSAR model, the following properties appear in the top-most equally helpful for other drug designing processes based on similar
equation: 3v cluster, 1j, Coulson charge on N3, and x component of s-triazine derivatives.
the dipole moment (lx). This list indicates that structural (topologi-
cal) as well as electronic factors contribute to the activity or inac-
tivity of a given compound. However, we require a deeper Acknowledgments
introspection of the actual quantitative effect of these parameters
on the activity value. Deciphering the information available from a One of the authors (P.G.) thanks the Council of Scientific and Indus-
QSAR model needs the study of coefficients of these properties as trial Research (CSIR), New Delhi, for a research fellowship. The
they appear in the top equations. authors thank University of Delhi's 'Scheme to Strengthen R&D Doc-
toral Research Program by Providing Funds to University Faculty'.
The most powerful factor here is the charge on the nitrogen atom
of the triazine ring, N3 sandwiched between the two side chains.
This indicates that the electron density at the triazine group should References
not decrease. So it is rational to attach electron donating atoms in
these side chains. 1. Hunt S.Y., Rezvani B.B., Sibley C.H. (2005) Novel alleles of Plas-
modium falciparum dhfr that confer resistance to chlorcyclogua-
3
v cluster contributes negatively toward activity. It leads us to keep nil. Mol Biochem Parasitol;139:25–32.
less clustering in the side chains. 1j has a positive coefficient, 2. Ferone R. (1977) Folate metabolism in malaria. Bull World Heath
although of a comparatively lower value. It signifies that contacts of Organ;55:291–298.
first degree between atoms are beneficial in improving the activity, or 3. Vilaivan T., Saesaengseerung N., Jarprung D., Kamchonwongpai-
we can say that branching is not a favorable trait. Clustering could san S., Sirawaraporn W., Yuthavong Y. (2003) Synthesis of
result in bad grades. Very long chains are also not recommended, as solution-phase combinatorial library of 4,6-diamino-1,2-dihydro-
elongation of the side chain has no major effect on the electronic con- 1,3,5-triazine and identification of new leads against
tribution toward activity. These points suggest choosing substituents A16V+S108T mutant dihydrofolate reductase of Plasmodium
consisting of simple 2-3 carbon atom chains to be substituted on the falciparum. Bioorg Med Chem;11:217–224.
s-triazine ring to improve the antimalarial activity. 4. Murray M.C., Perkins M.E. (1996) Chemotherapy of malaria.
Annu Rep Med Chem;31:141–150.
5. Agarwal A., Srivastava K., Puri S.K., Chauhan P.M.S. (2005) Syn-
Conclusions theses of 2, 4, 6-trisubstituted triazines as antimalarial agents.
Bioorg Med Chem Lett;15:531–533.
The results given above indicate that QSAR of minimum inhibitory 6. Srivastava S., Tewari S., Chauhan P.M.S., Puri S.K., Bhaduri A.P.,
concentrations of 2,4,6-trisubstituted-1,3,5-triazine compounds to Pandey V.C. (1999) Synthesis of bisquinolines and their in vitro
DHFR reductase can be modeled with a few molecular descriptors. ability to produce methemoglobin in canine hemolysate. Bioorg
The best model is a four-parameter regression equation with good Med Chem Lett;9:653–658.
statistical fit and good predictive power, as evident from its 7. Srivastava S., Tewari S., Srivastava S.K., Chauhan P.M.S., Bhad-
R2 = 0.7697, R 2CV ¼ 0:6469 values. An analysis of the descriptors uri A.P., Puri S.K., Pandey V.C. (1997) Synthesis of 7-chloro-4-
that are involved in the models indicates that the minimum inhibi- substituted aminoquinolines and their in vitro ability to produce
tory concentration is influenced largely by the Coulson charge on methemoglobin in canine hemolysate. Bioorg Med Chem
N3, which belongs to the 1,3,5-triazine ring, and is the nitrogen Lett;7:2741–2746.
atom sandwiched between the two substituted carbon atoms. 8. Baliani A., Bueno G.J., Stewart M.L., Yardley V., Brun R., Barrett
Another descriptor that influences significantly is 3v, which is an M.P., Gilbert I.H. (2005) Design and synthesis of a series of mel-
indicator of clustering in the compound. Other less significant de- amine-based nitroheterocycles with activity against trypanoso-
scriptors are 1j, and the x component of the total dipole moment. matid parasites. J Med Chem;48:5570–5579.
Thus, the analysis indicates good correlation between structure and

Chem Biol Drug Des 2011; 77: 57–62 61

Ojha et al.

9. Menicagli R., Samaritani S., Signore G., Vaglini F., Via L.D. 13. Jensen N.P., Ager A.L., Bliss R.A., Canfield C.J., Kotecka B.M.,
(2004) In vitro cytotoxic activities of 2-alkyl-4,6-diheteroalkyl- Rieckmann K.H., Terpinski J., Jacobus D.P. (2001) Phenoxyprop-
1,3,5-triazines: new molecules in anticancer research. J Med oxybiguanides, prodrugs of DHFR-inhibiting diaminotriazine
Chem;47:4649–4652. antimalarials. J Med Chem;44:3925–3931.
10. Henke B.R., Consler T.G., Go N., Hale R.L., Hohman D.R., Jones 14. Karelson M. (2000) Molecular Descriptors in QSAR ⁄ QSPR. New
S.A., Lu A.T. et al. (2002) A new series of estrogen receptor York, USA: Wiley Interscience.
modulators that display selectivity for estrogen receptor b. 15. Todeschini R., Consonni V. (2000) Handbook of Molecular
J Med Chem;45:5492–5505. Descriptors. Weinheim, Germany: Wiley-VCH.
11. Kuo G.-H., DeAngelis A., Emanuel S., Wang A., Zhang Y., Con- 16. Kier L.B., Hall L.H. (1976) Molecular Connectivity Indices In
nolly P.J., Chen X., Gruninger R.H., Rugg C., Fuentes-Pesquera Chemistry and Drug Research. New York: Academic Press.
A., Middleton S.A., Jolliffe L., Murraypp W.V. (2005) Synthesis 17. Dewar M.J.S., Zoebisch E.G., Healy E.F., Stewart J.J.P. (1985)
and identification of [1,3,5] triazine-pyridine biheteroaryl as a AM1: a new general purpose quantum mechanical method. J
novel series of potent cyclin-dependent kinase inhibitors. J Med Am Chem Soc;107:3902–3909.
Chem;48:4535–4546. 18. Friedman J.H. (1990) Technical Report No. 102. Laboratory for
12. Pandey V.K., Tusi S., Tusi Z., Joshi M., Bajpai S. (2004) Synthesis Computational Statistics, Stanford, California, USA: Department
and biological activity of substituted 2,4,6-s-triazines. Acta of Statistics, Stanford University.
Pharm;54:1–12.

62 Chem Biol Drug Des 2011; 77: 57–62