Molecular Descriptors Family on Structure Activity Relationships 5. Antimalarial Activity of 2,4-Diamino-6-Quinazoline Sulfonamide Derivates
 
Lorentz JÄNTSCHI and Sorana BOLBOACĂ
 
Technical University of Cluj-Napoca and “Iuliu Haţieganu” University of Medicine and Pharmacy, Cluj-Napoca, Romania;  http://lori.academicdirect.org   
 

 
Abstract
Antimalarial activity of sixteen 2,4-diamino-6-quinazoline sulfonamide derivates was modeled using an original methodology which assess the relationship between structure of compound and theirs activity. The results shows us that the antimalarial activity of studied 2,4-diamino-6-quinazoline sulfonamide compounds is alike topological and geometrical and is strongly dependent on partial change of the molecule. The ability in prediction with SAR models is sustained by the results obtained through cross-validation analysis and by the stability of the models. The SAR methodology gives us a real solution in structure-activity relationships investigation of 2,4-diamino-6-quinazoline sulfonamide compounds, obtaining better results by the use of two and/or three descriptors compared with the best performing previous reported model.
Keywords
Structure - Activity Relationships (SAR), Molecular Descriptors Family (MDF), Multiple Linear Regression (MLR), Antimalarial Activity, 2,4-diamino-6-quinazoline sulfonamide derivates
 
 
 
Background
 
            The sulfonamides are sulfa-related group of antibiotics used in bacterial and some fungal infections, killing the bacteria and fungi by interfering with cell metabolism. Sulfonamides and its derivates, including the 2,4-diamino-6-quinazoline sulfonamides [1], have been used in medicine for theirs antimalarial properties [2]. To date, for 2,4-diamino-6-quinazoline sulfonamide derivates, have been reported in specialty literature QSAR’s models using electronic parameters, as energy of highest occupied molecular orbitals (EH), energy of lowest unoccupied molecular orbital (EL) and charge density (CD) [3] and topological properties (Wiener index - W, Szeged index - Sz, and indicator parameters, called dummy or de novo constants, which take two values – zero or one – and serve as indication of category or class membership - Ip1, Ip2 and Ip3) [4]. Agrawal et all models the antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates by the use of topological properties and obtained mono-, bi-, tri-, and tetra-parametric models. The models obtained previously are in table 1, indicating the regression equations, the square of correlation coefficient (r2), and cross-validation values (r2cv) where were available.
 
Table 1. QSAR models for antimalarial activity of sulfonamide derivates reported by Agrawal
No
QSAR model
r2
r2cv
1
6.9977-0.0032( ±9.9221·10-4W
0.4229
-
2
6.8222-0.0019(±6.3548·10-4Sz
0.3713
-
3
9.6975-0.0045(±0.0010) ·W-1.9814(±0.08269)·Ip1
0.5997
0.3325
4
9.9246-0.0028(±6.9413·10-4Sz-2.1021(±0.8938)·Ip1
0.5590
-
5
9.4033-0.0041(±0.0010)·W-2.1844 (±0.8013)·Ip1
- 1.0922(±0.7280)·Ip2
0.6629
0.5009
6
9.4696-0.0026(±7.2234·10-4Sz-2.2182(±0.8888)·Ip1- 0.9272(±0.8072)·Ip2
0.6027
0.3343
7
9.0548-0.0019(±7.5770·10-4Sz-3.2559(±0.9977)·Ip1
-2.6109(±1.911)·Ip2-1.9345(±1.0718)·Ip3
0.6934
0.5579
8
9.1679-0.0032(±0.0010)·W-3.1824(±0.9143)·Ip1
-2.5978(±1.0591)·Ip2- 1.7911(±0.9807)·Ip3
0.7414
0.6493
 

            The aim of the research was to test the ability of SAR methodology in prediction of antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates and to compare the found models with previous reported QSARs.
 
 
Materials and Methods
 
Material and Pharmacology
            Sixteen 2,4-diamino-6-quinazoline sulfonamide derivates was included into analysis. The planar structure of 2,4-diamino-6-quinazoline sulfonamide derivates, the substituents X and Y, and the measured antimalarial activity (Yaa) are in table 2.  Antimalarial activity used in the study was taken from the paper reported by Elslager et all [1], and is defined as difference between the average survival times (in days) of treated mice and the average survival times (in days) of control mice.
 
Table 2. Planar structure of 2,4-diamino-6-quinazoline sulfonamide derivates and measured antimalarial activity

No
R1
R2
Yaa
mol_01
N(C2H5)2
H
3.3
mol_02
N(CH2)5
Cl
2.3
mol_03
N(CH2 CH2 CH3)2
H
0.3
mol_04
N(CH2 CH2 OH)2
H
0.3
mol_05
N(CH3)CH (CH3)2
H
0.7
mol_06
N(CH3)CH2CH2N(C2H5)2
H
0.1
mol_07
N(CH2)5
H
4.4
mol_08
N(CH2)4
H
5.0
mol_09
N[(CH2)2]2O
H
4.7
mol_10
N[(CH2)2]2S
H
2.5
mol_11
N[(CH2)2]2NCH3
H
1.0
mol_12
N[(CH2)]2NC(=O)OC2H5
H
0.2
mol_13
NH-C6H4-4Cl
H
0.7
mol_14
NH-C6H4-3Br
H
0.3
mol_15
NCH3-C6H4-4Cl
H
0.3
mol_16
NCH3-C6H5
H
0.3
 
SAR modeling
            The steps of molecular descriptors family on structure activity relationships modeling of antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates were [5]:
 
Results
 
            The best performing mono-, bi-, and tri-varied SAR models, together with associated statistics of regression analysis are in table 3.
 
Table 3. SARs for antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates with MDF members
No
SAR model
 
Characteristic
Notation and Value
1
Mono-varied model:             Ŷmono-v = 3.26·10-2 + 8.72·105·IsPmSQt
 
Correlation coefficient
r = 0.934
 
Squared correlation coefficient
r2 = 0.873
 
Adjusted squared correlation coefficient
r2adj = 0.864
 
Standard error of estimated
sest = 0.659
 
Fisher parameter
Fest = 96
 
Probability of wrong model
pest(%) = 1.2·10-5
 
t parameter for intercept; p-values
95% probability CIint [lower 95%; upper 95%]
tint = 0.140; ptint = 0.89
95%CI = [-0.467; 0.533]
 
t parameter for IsPmSQt descriptor; p for tIsPmSQt
95% probability CIIsPmSQt [lower 95%; upper 95%]
tIsPmSQt = 9.802; pIsPmSQt = 1.2·10-7
95%CIIsPmSQt = [6.81·105; 10.63·105]
 
Cross-validation leave-one-out (loo) score
r2cv-loo = 0.840
 
Fisher parameter for loo analysis
Fpred = 73
 
Probability of wrong model for loo analysis
ppred(%) = 6.2·10-7
 
Standard error for leave-one-out analysis
sloo = 0.741
 
The difference between r2 and r2cv(loo)
r2 - r2cv(loo) = 0.033
 
 
2
Bi-varied model:             Ŷbi-v = 4.81·10-3+1.95·105·IsMMEQt+2.27·107·IIMMTQt
 
Correlation coefficient
r  = 0.985
 
Squared correlation coefficient
r2 = 0.971
 
Adjusted squared correlation coefficient
r2adj = 0.967
 
Standard error of estimated
sest = 0.324
 
Fisher parameter
Fest = 220
 
Probability of wrong model
pest(%) = 9.4·10-9
 
t parameter for intercept; ptint
95% probability CIint [lower 95%; upper 95%]
tint = 0.039; ptint = 0.969
95%CIint = [-0.261; 0.271]
 
t parameter for IsMMEQt descriptor; pIsMMEQt
95% probability CIIsMMEQt [lower 95%; upper 95%]
tIsMMEQt = 7.702; pIsMMEQt = 3.4·10-6
95%CIIsMMEQt = [1.4·105; 2.5·105]
 
t parameter for IIMMTQt descriptor; pIIMMTQt
95% probability CIIIMMTQt [lower 95%; upper 95%]
tIIMMTQt = 17.74; pIIMMTQt = 1.7·10-10
95%CIIIMMTQt = [2·107; 2.5·107]
 
Cross-validation leave-one-out (loo) score
r2cv-loo = 0.961
 
Fisher parameter for loo analysis
Fpred = 163
 
Probability of wrong model for loo analysis
ppred(%) = 6.19·10-8
 
Standard error for leave-one-out analysis
sloo = 0.375
 
The difference between r2 and r2cv(loo)
r2 - r2cv(loo) = 0.00958
 
The squared correlation coefficient
between descriptor and measured
antimalarial activity, and between descriptors
r2(IsMMEQt, Yaa) = 0.277
r2(IIMMTQt, Yaa) = 0.840
r2(IsMMEQt, IIMMTQt) = 0.035
 
3
Tri-varied model:
Ŷtri-v = -17.6 + 6.83·108·IsMMTQt + 3.58·10-1·LsMrKQg -8.47·10-1·lsDMTQt
 
Correlation coefficient
r  = 0.998
 
Squared correlation coefficient
r2 = 0.997
 
Adjusted squared correlation coefficient
r2adj = 0.996
 
Standard error of estimated
sest = 0.1059
 
Fisher parameter
Fest = 1415
 
Probability of wrong model
pest(%) = 1.4·10-13
 
t parameter for intercept; ptint
95% probability CIint [lower 95%; upper 95%]
tint = -14.86; ptint = 4.32·10-9
95%CIint = [-20.23; -15.05]
 
t parameter for IsMMTQt descriptor; pIsMMTQt
95% probability CIIsMMTQt [lower 95%; upper 95%]
tIsMMTQt = 47.03; pIsMMTQt = 5.58·10-15
95%CIIsMMTQt = [6.5·108; 7.1·108]
 
t parameter for LsMrKQg descriptor; pLsMrKQg
95% probability CILsMrKQg [lower 95%; upper 95%]
tLsMrKQg  = 10.50; pLsMrKQg  = 2.09·10-7
95%CILsMrKQg = [0.28; 0.43]
 
t parameter for lsDMTQt descriptor; plsDMTQt
95% probability CIlsDMTQt [lower 95%; upper 95%]
tlsDMTQt = -17.07; plsDMTQt = 8.8·10-10
95%CIlsDMTQt = [-0.95; -0.74]
 
Cross-validation leave-one-out (loo) score
r2cv-loo = 0.9959
 
Fisher parameter for loo analysis
Fpred = 970
 
Probability of wrong model for loo analysis
ppred(%) = 1.4·10-12
 
Standard error for leave-one-out analysis
sloo = 0.1279
 
The difference between r2 and r2cv(loo)
r2 - r2cv(loo) = 0.0013
 
The squared correlation coefficient
between descriptor and measured
antimalarial activity, and between descriptors
r2(IsMMTQt, Ymaa) = 0.8448
r2(LsMrKQg, Ymaa) = 0.1556
r2(lsDMTQt, Ymaa) = 0.2493
r2(IsMMTQt, LsMrKQg) = 0.0135
r2(IsMMTQt, lsDMTQt) = 0.6140
r2(LsMrKQg, lsDMTQt) = 0.0242

            The list of descriptors and associated values used in mono-, bi-, and tri-varied models and estimated antimalarial activity (Ŷ) are in table 4.
Graphical representations of the antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates, obtained from structure for mono-, bi-, and tri-varied models vs. measured ones are in figures 1 to 3.
            Assessment of the MDF SAR model was performed by applying a correlated correlation analysis, which took into consideration mono-, bi-, and tri-varied SAR models and compared them with the best performing (model with four variables, r2 = 0.7414, r2cv = 0.6493) previous reported model [4] by the use of Steiger’s Z test. The results of comparison are in table 5.
 
Table 4. Descriptors used in MDF SAR models, theirs values and estimated antimalarial activities
 
Mono-varied
Bi-varied
Tri-varied
Mol
IsPmSQt
Ŷmono
IsMMEQt
IIMMTQt
Ŷbi
IsMMTQt
LsMrKQg
lsDMTQt
Ŷtri
mol_01
2.23·10-6
1.98
-2.01·10-6
1.53·10-7
3.07
7.63·10-9
-4.03·100
-2.02·101
3.24
mol_02
2.92·10-6
2.58
-1.46·10-7
6.98·10-8
1.56
3.17·10-9
-4.49·100
-2.28·101
2.23
mol_03
2.11·10-7
0.22
-1.12·10-6
1.57·10-8
0.14
7.13·10-10
-3.56·100
-2.21·101
0.29
mol_04
2.11·10-7
0.22
-1.12·10-6
1.57·10-8
0.14
7.13·10-10
-3.60·100
-2.21·101
0.28
mol_05
6.62·10-7
0.61
-1.28·10-6
6.60·10-8
1.25
3.30·10-9
-4.28·100
-2.08·101
0.73
mol_06
6.94·10-7
0.64
-6.59·10-7
1.94·10-8
0.32
8.08·10-10
-4.47·100
-2.22·101
0.14
mol_07
3.54·10-6
3.12
9.86·10-6
1.05·10-7
4.31
5.00·10-9
-1.45·100
-2.26·101
4.40
mol_08
5.91·10-6
5.19
4.82·10-6
1.73·10-7
4.87
8.65·10-9
-3.96·100
-2.14·101
4.95
mol_09
5.18·10-6
4.55
-1.76·10-6
2.31·10-7
4.89
1.10·10-8
-4.69·100
-1.96·101
4.76
mol_10
4.20·10-6
3.69
1.39·10-6
1.15·10-7
2.89
5.49·10-9
-4.37·100
-2.13·101
2.60
mol_11
7.38·10-7
0.68
-5.86·10-6
8.47·10-8
0.78
3.85·10-9
-5.24·100
-2.10·101
0.93
mol_12
1.17·10-7
0.13
-4.82·10-7
1.28·10-8
0.20
4.92·10-10
-4.41·100
-2.24·101
0.07
mol_13
1.02·10-6
0.93
-5.55·10-7
4.25·10-8
0.86
1.85·10-9
-4.68·100
-2.20·101
0.55
mol_14
1.86·10-7
0.20
-5.58·10-7
1.22·10-8
0.17
5.31·10-10
-5.03·100
-2.28·101
0.26
mol_15
7.33·10-7
0.67
-1.24·10-7
2.33·10-8
0.51
9.70·10-10
-4.13·100
-2.24·101
0.51
mol_16
1.12·10-6
1.01
-3.72·10-7
2.27·10-8
0.45
9.88·10-10
-5.09·100
-2.27·101
0.43

Figure 1. Measured antimalarial activity (MAA) vs. estimated (EAA) with mono-varied model


Figure 2. Measured antimalarial activity vs. estimated with bi-varied model



Figure 3. Measured antimalarial activity vs. estimated with tri-varied model

Table 5. The results of comparison obtained and best performing previous reported models
Characteristic
Values
Number of descriptors used in MDF SAR model
3
2
1
r(Yaa, ŶMDF SAR)
0.9986
0.9856
0.9342
r(Yaa, ŶPrevious)
0.8598
0.8598
0.8598
r(ŶMDF SAR , ŶPrevious)
0.8641
0.8624
0.8139
Steiger’s Z test parameter
7.8891
3.9686
1.3229
pSteiger’s Z (%)
1.5·10-13
3.6·10-3
9.2926

 
Discussions
 
            Antimalarial activity of sixteen 2,4-diamino-6-quinazoline sulfonamide derivates was modeled by the use of an original methodology which take into consideration the structure of the compound and try to explain the interest activity. Applying the MDF SAR methodology, three models, one mono-varied, one bi-varied and one-tri-varied prove to obtained performances in antimalarial activity prediction. All presented SAR models are statistically significant at a significance level less than 0.001. The mono-varied SAR model use a descriptor that take into consideration the topology of molecule (IsPmSQt) and the partial change as atomic property (IsPmSQt). Almost 87 percent of variation in antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates can be explainable by its linear relation with IsPmSQt descriptor. The mono-varied model is significant different by the best performing four-varied model previous reported at a significance level equal with 0.09. As the mono-varied SAR model, the bi-varied one took into consideration the topology of molecule (IsMMEQt, IIMMTQt) as well as partial change as atomic property (IsMMEQt, IIMMTQt). All coefficients of the bi-varied equation are significantly differed by zero, except the intercept of the slop. The performance of the bi-varied SAR model is sustained by the correlation coefficient and the squared of the correlation coefficient. Ninety-seven percent of variation in antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates can be explainable by its linear relation with IsMMEQt, IIMMTQt descriptors. The stability of the bi-varied model is proved by the very lower value of the differences between squared correlation coefficient and cross-validation leave-on-out squared correlation coefficient. The cross-validation leave-one-out score (r2cv-loo = 0.961) sustain the stability of the bi-varied SAR model. Looking at the values of the squared correlation coefficient between descriptors and measured antimalarial activity it can be observed that there is no correlation between IsMMEQt descriptor and measured antimalarial activity but there is a strong correlation between IIMMTQt descriptor and antimalarial activity. Even if the correlation is strong, the IIMMTQt is not the one that obtained best performances in terms of squared correlation coefficient and cross-validation leave-one-out score in mono-varied SAR model. It could not be observed a significant correlation between descriptors of the bi-varied model (r2(IsMMEQt, IIMMTQt) = 0.035). The bi-varied SAR model obtained a correlation coefficient significantly greater compared with the previous reported four-varied model at a significance level equal with 3.6·10-3 %. Note that, it is possible to obtained useful information about antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates with a bi-varied model instead of a model with four variable. Looking at the bi-varied model, we can say that the antimalarial activity is of molecular topology and depend on partial change of molecule. Looking at the cross-validation leave-one-out score, we can say that the tri-varied model is the best performing SAR model. Ninety-nine percent of variation in antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates can be explainable by its linear relation with  IsMMTQt, LsMrKQg, and lsDMTQt descriptors. Two descriptors (IsMMTQt, and lsDMTQt) take into consideration the topology of the molecule while another one (LsMrKQg) the molecular geometry. All three descriptors (IsMMTQt, LsMrKQg, lsDMTQt) take into consideration the partial change of the molecule. The values of squared correlation coefficient (r2 = 0.997) demonstrate the goodness of fit of the tri-varied MDF SAR model. The power of the tri-varied model in prediction of the antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide compounds is demonstrate by the cross-validation leave-one-out correlation score (r2cv(loo) = 0.9959), procedure which did not take into consideration one molecule from the whole set. The stability of the best performing tri-varied MDF SAR model is give by the difference between the squared correlation coefficient and the cross-validation leave-one-out correlation score (r2 - r2cv(loo) = 0.0013). Looking at the tri-varied SAR model we can say that the antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates is alike topological and geometrical and depend by the partial change of molecule. Looking at the correlated correlations analysis results, it can be observed that the tri-varied SAR model obtained a significantly greater correlation coefficient compared with the previous reported four-varied model, at a significance level equal with 1.5·10-13 %. Starting with the knowledge learned from the studied set of 2,4-diamino-6-quinazoline sulfonamide compounds, antimalarial activity of new compound from the same class can be predict by the use of an original software, which is available at the following address:
http://vl.academicdirect.org/molecular_topology/mdf_findings/sar/
           Thus, the software id able to predict the antimalarial activity of new 2,4-diamino-6-quinazoline sulfonamide compounds with low costs.
 
Acknowledgement
           Research was in part supported through project ET36/2005 by UEFISCSU Romania.
 
 
Conclusions
 
            Antimalarial activity of the studied 2,4-diamino-6-quinazoline sulfonamide compounds is alike to be by topological and geometrical nature and is strongly dependent by partial change. The MDF SAR methodology is a real solution in predicting antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide compounds and could be use in developing of new 2,4-diamino-6-quinazoline sulfonamide compounds with antimalarial properties. 
            Even if using of MDF in QSAR modeling is time consuming, it has doubtless advantages, such as better QSAR of antimalarial activity of 2,4-diamino-6-quinazoline sulfonamide derivates and a much closer structure activity explanation.
 
 
References
 

[1] Elslager E. F.,  Colbry N. L.,  Davoll J., Hutt M. P., Johnson J. L., Werbel L. M., Folate antagonists. 22. Antimalarial and antibacterial effects of 2,4-diamino-6-quinazolinesulfonamides, J. Med. Chem., 27(12), p. 1740-1743, 1984.
[2] Shao B. R., A review of antimalarial drug pyronaridine, Chin. Med. J. (Engl), 103, p. 428-434, 1990.
[3] Agrawal V. K., Sinha S., Bano S., Khadikar P. V., QSAR studies on antimalarial 2,4-diamino-6-quinazoline sulfonamides, Acta Microbiol Immunol Hung, 48(1), p. 17-26, 2001.
[4] Agrawal K. V., Srivastavaa R., Khadikarb V. P., QSAR Studies on Some Antimalarial Sulfonamides, Bioorgan. Med. Chem., 9, p. 3287-3293, 2001.
[5]  Jäntschi L., Molecular Descriptors Family on Structure Activity Relationships 1. The review of Methodology, Leonardo Electronic Journal of Practices and Technologies, AcademicDirect, 6, p. 76-98, 2005.
[6] ***, HyperChem, Molecular Modelling System [Internet page]; ©2003, Hypercube, Inc.,  available at: http://hyper.com/products/
[7] ***, Leave-one-out Analysis. ©2005, Virtual Library of Free Software, available at: http://vl.academicdirect.org/molecular_topology/mdf_findings/loo/