H-bond, – stacking and electronic interactions are shown as purple, orange and azure dashed lines, respectively
H-bond, – stacking and electronic interactions are shown as purple, orange and azure dashed lines, respectively. To look at Figure 7 as a whole, compound 13 inserts into the protein chain like an expansive V, which is composed of two (right and left) branches and one vertex corner. by using the Gasteiger-Hckel method , and then the conformer of each compound was energy-minimized using the Tripos molecular mechanics force field  and the Powell conjugate gradient minimization algorithm with the convergence criterion set to 0.05 kcalmol?1??1 to ensure the stability of the conformation. Since in 3D-QSAR studies, the most critical step is the molecular alignment , presently all 3D-QSAR statistical models were constructed based on the alignment of all the molecules, and compound 13 was chosen as the template due to GSK2801 its most potent activity in the data set. All the molecules were fitted into the template using the Align Database command in Sybyl. The common skeleton in the molecular superimposition is displayed in bold in Figure 1A,B depicts the resultant model. Open in a separate window Figure 1 Molecular alignments of all compounds in the data set. (A) The common structure of molecules based on template compound 13 is displayed in bold; (B) The resultant GSK2801 alignment model. 2.3. CoMFA and CoMSIA Studies To analyze the quantitative relationship between 3D structural features and the biological activity for a set of molecules, CoMFA and CoMSIA analyses were utilized for these antagonists after conformational alignment. All superimposed molecules were placed in a 3D lattice with spacing of 2.0 ?. CoMFA fields including the steric and electrostatic fields were generated by using sp3 C-atom probe with a formal charge of +1.0 at each lattice point and a van der Waals (vdW) radius of 1 1.52 ? . And both the steric and electrostatic fields were calculated by CoMFA standard method with energy cut-off values of 30.0 kcalmol?1 . CoMSIA is, though, an extension of CoMFA, it also includes extra hydrophobic, hydrogen bond (H-bond) donor and H-bond acceptor descriptors besides the steric and electrostatic descriptors. CoMSIA similarity index descriptors were derived by the same lattice boxes as those used in CoMFA calculations. And five different similarity descriptors were calculated by using a probe atom of charge +1.0, radius 1.0 ?. A Gaussian function was used to evaluate the GSK2801 mutual distance between each molecule atom and the probe atom, with no cut-off limits in CoMSIA study. In order to obtain statistically significant 3D-QSAR models and to analyze the relationship between their biological activities and the variations in CoMFA-CoMSIA interaction energies, partial least-squares (PLS) regression analyses were conducted [32,33]. PLS can reduce an originally large number of descriptors to some principal components which are linear combinations of the initial GSK2801 descriptors . In the present study, the CoMFA-CoMSIA descriptors were used as independent variables, while dependent variables were the pIC50 values. In PLS analysis, the leave one out (LOO) method that one molecule is removed from the data set and its activity is predicted by a model derived from the remainder of the data set, was used to evaluate the reliability of model by calculating the conventional correlation coefficient (value were calculated . = but low SEE values are also expected for a reliable QSAR model . In the present work, by means of PLS statistical analysis, the resultant CoMFA model obtained by using both the steric and electrostatic field descriptors is unsatisfied with value of 94.879 and a low SEE value of 0.161 with 7 optimum components, implying a good internal predictability. And the relative contributions of the steric, electrostatic, H-bond donor and H-bond acceptor fields are 24.5%, 45.4%, 14.9% and 15.2% in turn. The higher contribution of the electrostatic field indicates that electrostatic feature plays more roles in the antergic activity for the series. Some models are found to have admissible internal predictability but unfavorable external predictability . Thus, the em R /em 2pre should also be considered for a reliable model. In this study, a test set of 27 compounds, representing 32.9% of the training set, was employed to validate the robustness of the models. In general, the em R /em 2pre Rabbit polyclonal to GNMT above 0.6 is an acceptable standard . The developed CoMSIA model gives a high em R /em 2pre value of 0.897, much higher than this criterion. Therefore, the CoMSIA model is selected as the optimal one, and its correlation for the whole dataset is described in both scatter plot (Figure 3) and radar plots (Figure 4). As we can see, all the data points distribute uniformly around the regression line in Figure 3, and the data lines overlap within the low deviation in both Figure 4A,B, which illustrate the good correlation of the predicted bioactivity data versus the experimental data, as well as.