The data we used was taken from the paper [16] . The structures of chalcone derivatives used in this study were listed in Table 1 and their biological activity IC₅₀ (half maximal inhibitory concentration) values µM (for inhibitor of the human colon cancer). The IC₅₀ was used as a dependent variable in the QSAR study and all original IC₅₀ values were converted to negative logarithm of IC₅₀ (pIC₅₀). These data were importantly used to establish a 3D-QSAR model to observe the structure-activity relationship of chalcone derivatives.

Molecular similarity searching is fast becoming a key tool in drug design. CoMFA is a method that reflects the non-bonding interaction between the receptor and the ligand [17] . CoMFA, analyses the relationship of compounds’ structure and activity, is widely used in drug design. The results of statistical analysis can be graphically outputted on the molecular surface to indicate steric and electrostatic properties, and can help understand the detail of interaction between the ligand and the active site of recepotir [18] .

First, molecular alignment is an important step in CoMFA operation, directly affects the predictive ability of the CoMFA model. In this study, 20 chalcone derivatives were used as the training sets and 4 chalcone derivatives were used as the test sets. The data sets were used to construct 3D-QSAR model and analysis their physicochemical properties. The most active chalcone compound 21 was selected as a common skeleton which was selected for the alignment of training molecules. As shown in Figure 1, it’s compound 21 and shows the selected common skeleton. Figure 2 shows the result of compounds of training sets aligned.

Table 1. Chalcone derivatives and their experimental and predicted IC₅₀.

*test data sets.

Figure 3 shows the correlation between the experimental value from the literature [16] and the predicted value of chalcone CoMFA model. As shown in the picture, the abscissa represents the experimental value, the ordinate represents the predicted value, the black points represents the training set and the red points represents the test set. And all the samples are concentrated near the line y = x, which indicates that the model has better fitting ability and predictive ability and the predicted value of model was similar to the experimental value.

The countour map of chalcone is used to represent the steric and electrostatic fields of 24 compounds in different colors, as shown in Figure 4. Figure 4(a) shows steric interactions by green and yellow colored contours. The green regions indicate that the bulky substituents are favored, and yellow regions increase activity with small volume substituents. Figure 4(b) shows electrostatic interactions by red and blue colored contours. The blue regions indicate that positive charges are favored, and red regions increase activity only with negative charges.

There is major green region around R₂ position in the steric contour map suggesting that a bulky group is preferred in the position to produce higher

Q².: Cross-validated correlation coefficient. ONC.: Optimum number of components. R²: Non-cross- validated correlation coefficient. S_cv.: Standard error of the estimate. F: test value.

antitumor activity. For example, 9 compounds (3 - 11) have higher antitumor activity than compound 2, because a mono-halogen and mono-methyl groups have bulky volume than hydrogen. An analysis that a relatively bulky substituent is beneficial to the antitumor activity. The electrostatic contour map shows a blue region surrounding the R₂ position, indicating that substituting which can increase the positive charge on the ring systems would result in better activity. It is found that an electron-withdrawing group like a mono-halogen increases the antitumor activity, such as compounds (3 - 8, 17 - 19). An analysis that a relatively electron-withdrawing substituent is beneficial to the antitumor activity.

According to the contributions of the steric and electrostatic fields in the CoMFA model, the contribution of the steric field was greater than the contribution of electrostatic field. According to the analysis of the contour map, a relatively bulky substituent and electron-withdrawing substituent around R₂ position could increase the antitumor activity of chalcones. Then some new chalcone molecules were designed, as shown in Table 3.

Table 3. Structure and predicted pIC₅₀ values of newly designed drug molecules with antitumor.

The pIC₅₀ value is larger, the antitumor activity of compounds is better. Finally, CoMFA model was used to predict the antitumor activity of new chalcones, then we found that the pIC₅₀ values of the designed structures (N1, N2, N4, N6) were 6.090, 6.101, 6.050 and 6.064 respectively, were better than other compounds’ pIC₅₀, because these pIC₅₀ were bigger than the pIC₅₀ of some compounds in Table 1. The results proved that this model can provide useful indicators for further design of new drug.