QSAR treatment of meisoindigo derivatives as a potentbreast anticancer agent

A quantitative structure-activity relationship (QSAR) analysis of meisoindigo derivatives as a breast anticancer has been carried out. This study aimed to obtain the best QSAR model in order to design new meisoindigo based compounds with best anticancer activity. The semiempirical PM3 method was used for descriptor calculation. The best QSAR model was built using multilinear regression (MLR) with enter method. It was found that there were 19 new meisoindigo derivativeswith better predictive a potent anticancer agent. The best compound was (E)-2-(1-((3-ethylisoxazol-5-yl)methyl)-2-oxoindolin-3ylidene)-N-(4-methoxyphenyl)acetamide with the value of IC505.31144 x10 -15 (μM).


INTRODUCTION
Cancer is an abnormal growth of cells in the body's tissues which gradually become malignant. These cells grow faster and cannot experience apoptosis. Based on the data reported in 2016 there were around 14 million cancer cases in which 8.2 million cases were reported dead [6]. The type of cancer that causes most death in women, especially those aged 40 years and over, is breast cancer. Experimental research and development of new drugs in the laboratory requires considerable time and costs. In addition, the results obtained are likely to be unsatisfactory, so that a series of laboratory work becomes useless and wasted. Application of computational chemistry can be important in medicinal chemistry, especially for drug design, chemical properties theory, and biological activity of a molecule [8]. One of the potential compounds for anticancer is meisoindigo derivative compound. The in-vitro test for breast cancer cells shows that meisoindigo derivative compounds have good activity against these cancer cells [4].
This study will carry out a QSAR analysis of meisoindigo and its derivative compounds with the parameterized Model 3 (PM3) semiempirical method. The results of this study are the QSAR equation, which will then be used as a reference for the design of new meisoindigo derivative compounds as anticancer for abreasts cancer cells that have better biological activity.

Molecular Modeling.
Modeling of the molecular structure of the meisoindigo derivative compound was carried out using the PM3 semiempirical method and with a Convergence Limit of 0.001 kcal / Å. Optimization was carried out based on the Polak-Ribiere algorithm to achieve the minimum energy with Root Mean Square (RMS) of 0.001 kcal / Å [1]. Descriptors that will be counted represent 3 main parameters including hydrophobic, electronic, and steric parameters. The materials used were 20 types of meisoindigo derivative compounds obtained from the results of the study [4] which were known to have inhibitory activity on CDK4. The structure of the parent compound from the meisoindigo derivative can be seen in Figure 1. Furthermore, for the data of compounds with biological activities, see Table 1 . QSAR equation analysis was done by statistical analysis using the SPSS enter method, which was conducted by entering several selected independent variables (data descriptors), that generated some equation models which were the relation of physico-chemical properties of independent variables with anticancer activities (Log 1/IC 50 ) as a dependent variable. The models are then tested for validity [3].
Validation of QSAR Equations. The chosen QSAR model must meet all statistical criteria used and had specific descriptors that represented hydrophobic, electronic, and steric parameters. Validity testing of several equation models wasdone by calculating the value of r (correlation coefficient) and r 2 (coefficient of determination) close to 1, adjusted r 2 with the biggestvalue, smallest SE (standard error), F count /F table > 1 and PRESS (predicted recidual sums of squares) with the smallest value [10].
New Compound Design. The design of the new compound was done by modifying the type and position of the substituent. The position of the substituent was focused on the active site. The active site was chosen because it has atoms that are responsible for a breast anticancer activity from meisoindigo derivative compounds. The next step was to perform calculations using the semiempirical PM3 method according to the method used in determining the QSAR compound against all new compounds. Theoretical breast anticancer activity, Log 1/IC 50 from the yield compound, was calculated using the QSAR equation that hadbeen selected. Compounds with a high Log 1/IC 50 value indicate that these compounds have high at breast anticancer activity and these compounds can be proposed for synthesis [2].

RESULTS AND DISCUSSION
Descriptors used in this study are descriptors representing electronic parameters, hydrophobic and steric. The electronic parameter consists of the net charge of the atom (q), isolated atomic energy (Eat.is), dipole (μ), hydration energy (EH), binding energy (Eb). Descriptors representing hydrophobic parameters are partition coefficient of n-octanol / water (Log P), meanwhile descriptors representing steric parameters consist of the gradient (Gr) and surface area approx (SA). The net charge of atoms used is atoms C6, C8, C9, C10, N11, C13, O14, and C15. Approx (SA) surface area descriptors, hydration energy (EH), and n-octanol / water (coefficient P) partition coefficients were obtained from the structure of the compound optimized in Hyperchem 8.0 software by selecting the next compute menu then selecting QSAR properties. Whereas for the atomic net charge (q), gradient (Gr), isolated atomic energy (Eat.is), dipole (μ), and binding energy (Eb) descriptors obtained from the saved record file in the form (file.log).
The QSAR equation can be obtained from various multivariate statistical methods that produce satisfactory results. The most widely used basic method is regression analysis. The method correlates several independent variables x (in the form of physical-chemical parameters in the Hansch method or the value of indicator variables in the Free-Wilson method) with nonindependent or bound variables y (in the form of biological activity parameters of the compound). In this study, the QSAR equation was determined by regression analysis. The analysis is done by selecting the Analysis menu, then regression, and finally select linear. Determine the dependent variable, which is an experimental Log 1/IC 50 value, and an independent variable, which is 16 descriptors. The next step selects the enter method and clicks ok. The results of multilinear regression analysis are presented in table 3 for 14 equation models. 2,5 S11* 6,2 S16 3,1 S14 13,4 S5* 3,2 S1* 14,0 S2 3,5 S15 27,9 S17 3,5 S19* 44,7 S4 3,6 S20 64,3 S13 3,9 S18 84,9 Compounds with a sign (*) are compounds that are used as a test set.
The selected QSAR equation model must meet all the statistical criteria used and have specific descriptors that represent each hydrophobic, electronic and steric parameter. The statistical criteria used are the value of the correlation coefficient (r), the coefficient of determination (r 2 ), adjusted r 2 , a standard of error (SE), and F count /F indicates that model 14 has a good accuracy in describing the relationship between independent and non-independent variables. This is also supported by the standard of error data where model 14 has a relatively smaller error compared to other models. The other statistical criteria are F count /F tabel . The F count /F tabel value is able to describe the significance level of the data with a confidence level of 95%. The greater the F count /F tabel value, the more significant the data will be. Since the 6-12 model has a value of F count /F tabel > 1, it is assumed that all of these models are statistically acceptable. While models 4, 5, 13 and 14 have F count /F tabel values close to 1 which indicates that the models are close to significant. Based on the analysis of statistical criteria from r; r 2 ; adjusted r 2 ; SE; F count /F tabel ; and the PRESS test, the models 4, 13 and 14 are considered to be the best QSAR anticancer equation model of meisoindigo derivative compounds.
The selection of the best QSAR models from the three models was carried out by validation tests using four predetermined test sets. Validation test was done by predicting a breast anticancer activity using three selected equation models, then making a log1/IC 50 relationship curve experiment with prediction 1/IC 50 as seen in table 4 and in figure 2.
The relationship between the comparison of Log 1/IC 50 experiments with predictive Log 1/IC 50 will produce slope values and r 2 values that can be taken into consideration in determining the best equation model. If the predicted value of Log 1/IC 50 approaches 1/IC 50 experiment, the slope value on the curve will get closer to one. Therefore to determine the QSAR equation model is best done by finding the slope value that is closest to one. Based on the comparison curve between predictions and experiments, it can be seen that the slope value that is closest to one is model 14, with a value of 0.821. While for models 4 and 13 give a slope value of 0.411 and 0.632, respectively.
The value of r 2 which is close to 1 indicates that the correlation between the independent variable and the dependent variable is significant but if the value of r 2 approaches 0 then the correlation between these variables is not significant. Based on the curve, it can be seen that the value of r 2 which is closest to one is model 14, which is 0.      There are several factors for selecting the proposed compound for analysis besides paying attention to the IC 50 value which is of small value. Compound number 25 forms a structure similar to compound 8 which experimentally has the best activity. Based on its substituents, the methoxy group (p-MeO-Ph) in R3 is a group that is an electron withdrawing that is stronger than hydrogen by showing a negative induction effect, so it is used to increase electronic parameters that play a role in drug-receptor interactions that can later provide a biological response. The substitution of the methoxy group is carried out in the -para position in order to influence the steric properties of the drug molecules so that it can prolong the action of the drug in the body [11].

CONCLUSIONS
The best QSAR equation model of meisoindigo derivative as anticancer in breast cancer using semiempirical PM3 method is as follows: