Abstract: This report describes QSAR and SAR studies on the Inhibition of Reveres Transcriptase (RT) by 79 TIBO (Tetrahydoimidazobenzodizepin-2-one) derivatives using both classical and unconventional physicochemical properties and quantum molecular descriptors along with indicator parameters. The application of a multiple linear regression analysis indicated that a combination of classical physicochemical descriptors and the indicator parameters yielded a s tatistically significant model for the activity, log 1/C (50% of inhibition concentration of TIBO derivatives for RTs). The final selection of a potential TIBO compound for the inhibition of Reveres Transcriptase is made by quantum molecular modeling. We have found that, among the a number of Quantum and modeling parameters, the electron density on the 9th atom correlated best with the activity.
INTRODUCTION
The Acquired Immune-Deficiency Syndrome (AIDS) is one of the most hazardous diseases, which is caused by infection with the Human Immunodeficiency Virus (HIV). Reverse Transcriptase (RT) is the key for HIV replication and is not required for normal host cell replication. During the inspection of effective therapies facing HIV, Reverse Transcriptase (RT) has been identified as one of the most promising targets (Garg et al., 1999).
Substituted TIBO (tetrahydoimidazobenzodizepin-2-one) derivatives find
extensive applications in inhibition of reverse transcriptase (Garg et
al., 1999). It is, thus, important to understand the potential inhibition
activity, electronic and steric features. Thus, there is a compelling
need to understand the mechanisms and correlation modes of potential inhibition
activity with reverse transcriptase. Over the past several decades, the
Hansch group has studied the effect and significance of the molecular,
steric and hydrophobic parameters in the modeling of a number of biological
activities of numerous compounds. In a review based on the QSAR study
on NNRTI (non-nucleoside reverse transcriptase inhibitors), the Hansch
group (Garg et al., 1999) has reported that the use of steric and
molecular parameters yielded excellent statistics on the large set of
TIBO derivatives. Inspired by the pioneering work of Hansch, (Garg et
al., 1999) and in the continuation of our earlier works, (Wlodawer,
2002; Barre et al., 1983; Gallo, 1984; De Clercq, 1992; Mitsuya
and Broder, 1986; Reardon and Miller, 1990; Richman et al., 1987;
Fischl et al., 1989; Debyser et al., 1992) we have revisited
the work of Hansch (Garg et al., 1999) to see if we can further
develop a significant QSAR model with an entirely different set of parameters.
| Fig. 1: | Parent structure of TIBO derivative used in present study |
To achieve this objective, we have used a large set of Molecular descriptors such as Molar Refractivity (MR), Molar Volume (MV), Parachore (Pc), the Refractive index (ç), Surface Tension (ST), Density (D), Hydration Energy (HE), Approximate Surface Area (ASA), Surface Area Grid (SAG) and quantum chemical parameters, along with the indicator parameters used for the structural and positional specifications. To obtain a statistically significant model, we have used a number of methods including the maximum R2 method, which was followed by stepwise regression analyses (Loya et al., 1992, 1994, 1995a, b; Chaterjee and Hadi, 2000).
We note that the maximum R2 method actually includes a combination of standard error, adjusted R2 value (RA2), R, standard error of estimation and the F-ratio value. The predictive ability of the model is discussed on the basis of the predictive correlation coefficients. To validate our model further, we have used quantum molecular modeling parameters and on the basis of these parameters, we have analyzed the structural behavior of these molecules. For the molecular modeling, we have optimized the geometries of molecules using the molecular mechanics method by applying the MM+ force field method and Huckel molecular orbital theory using CNDO (Complete neglect of differential overlap) methods. Our motivation for the Huckel molecular orbital theory study is that the electronic and quantum parameters strongly depend on the degree of sophistication and such electronic parameters vary with the degree of the level of theory. The accuracy of a molecular mechanics or quantum mechanical method depends on the database used to parameterize the method.
MATERIALS AND METHODS
Activity
The log 1/C (IC50) of the TIBO derivatives (Fig. 1) is
adopted from the literature (Garg et al., 1999).
Physicochemical Parameters
The Molar volume, Parachor, Molar Refractivity, Refractive Index,
Surface Tension, Density and PolarI Zability, for the set of TIBO derivatives
were calculated from ACD Lab software (http://www.acdlabs.com)
and the un-conventional physicochemical parameters Approximate Surface
area, Surface area grid and Hydration Energy were calculated using hyperchem
7 demo-version. Molecular modeling parameters were calculated by applying
MM+ force field (Molecular mechanics) using Hyperchem 7 demo-version
(http://www.hyper.com).
Indicator Parameters
Indicator parameters are the dummy parameters sometimes used for accounting
those structural feature not covered in any molecular descriptor used.
They assumed only two values 1 or 0. If the assumed structural feature
is present; then the indicator parameters are 1 otherwise it is 0. The
details of such parameters, used in the present study are already given
in the Result and Discussion section (Table 2).
Statistical Analysis
Maximum R2 method together with stepwise regression (Chaterjee
and Hadi, 2000) was carried for arriving at statistically significant models.
In present study linear mathematical models are developed to study Quantitative
Structure/Property-Activity Relationship. Multiple linear regression analysis
is used to develop these models.
RESULTS AND DISCUSSION
The set of 79 TIBO derivatives and their adopted activity i.e., the inhibition activity of Reverse transcriptase-1 had been expressed as log 1/C is shown in Table 1.A very low-level degeneracy is present in the activity log1/C (mol L-1). As a result of the occurrence of degeneracy in activity log1/C, it becomes essential to examine the degeneracy in the molecular descriptors also (Table 1). A perusal of Table 2 and 3, which contains the unconventional physicochemical parameters and classical physicochemical parameters calculated for TIBO derivatives, shows that the low level degeneracy is observed in the unconventional and classical physicochemical descriptors (Balaban, 1992; Balaban and Balaban, 1991) has shown that the indices/descriptors, in spite of their degeneracy, can be used successfully in developing statistically significant QSAR models.
The correlation among the descriptors like unconventional physicochemical parameters (Table 2), classical physicochemical properties (Table 3), indicator parameters (Table 2), logP (Table 2) and activities shows (Table 4) that, except for the MR, logP, SAG and indicator parameters IZ and IR, all other unconventional and classical physicochemical parameters do not correlate well with the biological activity log1/C.
Out of the set of unconventional physicochemical descriptors used, Initial
bi-parametric regression analyses indicate that the combination of surface
area grid and the indicator parameter IZ plays the significant
role in modeling the activity log1/C. But the statistics obtained from
this combination is not adequate to explain the structure activity relationship.
In the case of trivariate correlations, the combination of SAG, IZ,
IR resulted little better but not as required. In case of tetra
and penta variate correlations the results are encouraging and the best
results obtained from the tetra variate combination of SAG, IZ,
IR, IX with the biological activity log1/C. The
model obtained from the above variables is:
| Table 1: | Substituents and biological activity of TIBO derivatives used in present study |
| Table 2: | Unconventional physicochemical parameters, logP and indicator parameters used in present study |
| ASA = Approximate Surface Area (A°2), SAG = Surface Area Grid (A°2), HE = Hydration Energy (kcal mol-1), logP = Hydrophobic parameter (Octanol/water Partition coefficient) IZ = 1 if S atom at Z position, IR = 1 if Acyclic structure at R position, IX = 1 if halogens present at X position | |
| Table 3: | Classical physicochemical properties used in present study for TIBO derivatives |
| MR = Molar Refractivity (cm3), MV = Molar Volume (cm3), Pc = Parachore (cm3), η = Index of Refraction, ST = Surface Tension (dynes cm-1), D = Density(g cm-3), α | |
|
log1/C=0.0052(±0.0036)SAG + 1.2207(±0.2176)IZS |
(1) |
| n = 79,Se = 0.8158,R = 0.8407,R2A = 0.6907, F = 44.591 |
The model obtained from the above combination is below:
|
log1/C=0.0389(±0.0186)MR + 1.1267(±0.2231)IZ |
(2) |
| Table 4: | Correlation matrix between unconventional, classical physicochemical parameters, logP, indicator parameters and log1/C |
| n = 79,e = 0.8037,S R = 0.8458, R2A = 0.7001,F = 46.516 |
Same procedure was followed for the estimation of log1/C from the logP
and the best model obtained from above variables is:
log1/C=0.2713(±0.126) logP + 1.3211(±0.2014) IZS |
(3) |
| n = 79, Se = 0.8023,R = 0.8464, R2A = 0.7011,F = 46.732 |
Substitutional effects are shown by the indicator parameters. Correlation matrix (Table 5-7) shows that all three indicator parameters IZ, IR and IX are having good correlation coefficients (0.684, 0.660 and 0.576, respectively) individually with biological activity log1/C.
Equations suggest that the positive correlation coefficient of unconventional physicochemical parameter surface area grid, classical physicochemical property MR show direct relationship with biological activity log1/C. The hydrophobic parameter logP also shows positive correlation coefficient bears direct relationship with biological activity. The positive correlation coefficient of indicators IZ, IR and IX also shows positive impact on the biological activity quantitatively.
In view of this above, we have concentrated on the results given by Eq. 3. Further regression analysis indicated that the model expressed by Eq. 3 has nineteen outliers in five different steps (compounds 5, 22, 24, 28, 29, 30, 32, 34, 45, 52, 60, 67, 68, 70, 72, 73, 75, 76 and 79), the deletion of which give the following models with excellent statistics:
log1/C= 0.2269(±0.1106)logP + 1.2569(±0.1739) IZS |
(4) |
| Fig. 2: | Relationship obtained between observed and calculated log1/C from Eq. 3 |
| Fig. 3: | Relationship obtained between observed and calculated log1/C from Eq. 8. |
| n = 74,Se = 0.6570,R = 0.9025,R2A = 0.8038,F = 75.757 |
|
log1/C=0.2240(±0.0939)logP + 1.2700(±0.1515) IZ |
(5) |
| n = 69,Se = 0.5542, R = 0.9326,R2A = 0.8616,F = 106.863 |
log1/C=0.2603(±0.0823)logP + 1.2883(±0.1381) IZ |
(6) |
| n = 66,Se = 0.4833, R = 0.9505,R2A = 0.8971,F = 142.709 |
log1/C = 0.2784(±0.0673)logP
+ 1.2539(±0.1154) IZS |
(7) |
| n = 61, Se = 0.3931, R = 0.9697, R2A = 0.9404, F = 220.894 |
log1/C = 0.2820(±0.0634)logP + 1.2077(±0.1033)
IZ |
(8) |
| n = 60, Se = 0.3703, R = 0.9736, R2A = 0.9440, F = 249.77 |
| Table 5: | Observed and calculated log1/C of TIBO derivatives used in present study |
| a: Calculated log1/C values from Eq. 3. b: Calculated log1/C values from Eq. 8, *: Data point not included in calculations from Eq. 8 | |
| Table 6: | Modeling parameters calculated for the few TIBO derivatives |
TE = Total Energy, DpM = Dipole Moment, RMSg = Root Mean Square gradient |
|
| Table 7: | Modeling parameters* calculated for the few TIBO derivatives |
| NC2 = Net charge on 2nd Carbon atom, NC6 = Net charge on 6th Nitrogen atom, NC9 = Net charge on 9th Carbon atom, ED2 = Electron Density on 2nd C atom, ED6 = Electron Density on 6th N atom, ED9 = Electron Density on 9th C atom (All in a.u.) | |
At this stage, it is worthy to comment on R2A values.
We observed that as we pass from the model obtained for 79 compounds (Eq.
1-3) to model obtained for 60 compounds (Eq. 8)
there is consistent increase in R2A, increasing
from 0.7011 to 0.9440, as we pass from (Eq. 3-8).
Such an increase in R2A values indicates that the
deleted compounds have the unfair share in the modeling of respective
activity and also showing exceptional behavior from their parent series.
The value of R2A will decrease if the deletion of
the compounds does not reduce the unexplained variation in the model enough
to off set the loss of degree of freedom (Lawtrakul et al., 1999;
Tanaka, 1991, 1992, 1995). In the second phase of our study based on the
category second containing 19 compounds. All these compounds (outliers)
taken together resulted into a model according to the following equation:
|
log1/C = 22.2389(±5.9919)
η + 7.5472 (±3.7888) J-1.3837(±0.6223) IR-43.166 |
(9) |
| n = 19, Se = 0.8207, R = 0.7422, R2A = 0.461, F = 6.131 |
Model presented in form of Eq. 9 expresses the domination of steric and structural features in comparison with hydrophobic parameter logP for the set of 19 out lairs.
Equation 9 also shows the domination of steric properties over the branching and size specific properties for modeling the activity log1/C for the set of 19 outliers. Equation also demonstrates the high favor of branching and steric property to the biological activity log1/C. Equation also exhibits the unfavorable presence of acyclic structure at R position for these compounds, just apposite to the parent series.
We have also obtained quantum chemically derived parameters since some of the properties depend strongly on electronic features such as electrophilic regions of the compounds. In order to carry out quantum computations, we have first carried out the molecular geometry optimizations (Loya et al., 1995) to find out the structural behavior of these compounds as a function of attached groups and their positions. The corresponding molecular modeling parameters are shown in Table 6 and 7.
Based on the above study and magnitude of residue from Eq. 8 we have selected compounds viz., 25, 39, 41, 43, 50, 55, 65, 71 and 77 to correlate their modeling parameters with the activities. This we have done to find out which TIBO derivative has the highest correlative and predictive potential for the same category. The molecular modeling is demonstrated in (Fig. 4-12), respectively for compounds 25, 39, 41, 43, 50, 55, 65, 71 and 77. The corresponding molecular modeling parameters are presented in Table 5. In order to resolve our problem of selecting out the TIBO derivative with the best quality and correlation potential; we have carried out further regression analysis using the molecular modeling parameters from Table 6 and 7.
From the modeling parameters significant univariate correlation shown
by the net charge on atom 2 (C), electron density at atom 2 (C), electron
density on 9th atom, but excellent result obtained from Net charge on
9th atom (C) and the models obtained are shown below.
log1/C = -9.9796 (±
2.8803) NC2 + 9.7946 |
(10) |
| n = 9, Se = 0.9703, R = -0.7948, F = 12.004 |
log1/C = 9.9775 (± 2.8803)
ED2 -30.1140 |
(11) |
| n = 9, Se = 0.9705, R = 0.7947, F = 11.999 |
|
log1/C = -65.0924 (± 13.8025)
ED9 -255.2617 |
(12) |
| n = 9, Se = 0.7823, R = -0.8723, F = 22.24 |
|
log1/C = 80.6404 (± 12.3788)
NC9 -7.7073 |
(13) |
|
n = 9, Se = 0.6016, R = 0.9265, F = 42.437 |
| Fig. 4: | Opt. Structure of Comp. 25 |
| Fig. 5: | Opt. Structure of Comp. 39 |
| Fig. 6: | Opt. Structure of Comp. 41 |
| Fig. 7: | Opt. Structure of Comp. 43 |
| Fig. 8: | Opt. Structure of Comp. 50 |
| Fig. 9: | Opt. Structure of Comp. 55 |
| Fig. 10: | Opt. Structure of Comp. 65 |
| Fig. 11: | Opt. Structure of Comp. 71 |
| Fig. 12: | Opt. Structure of Comp. 77 |
Equation 10 demonstrates that the compound having the carbon atom at 2nd position with higher net charge is unfavorable for the inhibition activity against RT`s for the TIBO derivatives. At the same time opposite results are shown by the electron density at the same carbon atom in Eq. 11. Both the equations exhibit the significant role of the nature of the C atom at 2nd position in the inhibition of RT1 by TIBO derivatives.
The results obtained from Eq. 12 and 13 express the unfavorable presence of the electron density on carbon atom at 9th position and the presence of the high net charge, at the carbon atom on 9th position in positive manner for the inhibition of RT1 by TIBO derivatives. The comparison of all the four equations exhibit the domination and significant role of 9th atom (C) over the atom 2 (C) i.e., the presence of any substitution at C atom at 9th position containing high net charge favors the inhibition activity for RT1 by TIBO derivatives as compare to substitution containing low electron density at the same atom as well as the substitution with low net charge and high electron density at 2nd carbon atom.
CONCLUSION
From the result and discussion made above, we conclude that the hydrophobic parameters can be used successfully for modeling the inhibition activities of reverse transcriptase-1 by TIBO derivatives and that for the present set of TIBO derivatives the hydrophobic parameter logP is found to be the prominent one. The results also indicate that combination of unconventional, classical and hydrophobic parameters and molecular (3D) modeling can be used for the understanding the structural behavior and select the compound with potential activity.