SAR of Triazole-Derivatives

The data set consists of 23 triazole-derivatives [148]. The molecular formula investigated of these compounds is shown in Fig. 9.4. According to the bioactivities MIC (Candida albicans931103) of samples, the data set can be divided into two classes. Herein samples of class 1 denote the compounds with high activity, i.e., the molecules with MIC 80jWg ml, while samples of class 2 denote the compounds with low activity, i.e., the molecules with MIC 80//g ml.  

In view of the complexify of the molecules studied, the method of calculation is simplified in the following way a series of sub-structures with different substituted group R (right part in Fig. 9.7) are used for the computation of molecular mechanics and quantum chemical methods. This kind of simplification is similar to the process of Hansch method considering the effects of various substituted groups. The difference of our method with Hansch method is that we did not treat the contribution of atomic groups separately, but treat the every substructure as a whole 

A SVC-based feature selection method is applied for descriptor selection. Using floating search technique, the subset of descriptors is selected out via LOO cross-validation. By this way, three descriptors, Cn (charge density of the atom N connecting with R), Dipole (dipole moment) and Hf (heat of formation) are selected for modeling use. Table 9.4 lists the samples with bioactivities MIC and selected descriptors. 

Release 7.0 for Windows Molecular Modeling System, Hypercube Inc. 2002 

5 Selection of the kernel function and parameter C used in SVC model 

---