Novel 3D Descriptors Using Excluded Volume 2: Application to Drug Classification
Yukio Tominaga · Journal of Chemical Information and Computer Sciences · 1998
We have developed novel 3D descriptors, van der Waals excluded volume of each molecule and probe, which do not require alignment rules, and are not significantly affected by the orientation of molecules. Each probe is constructed by the excluded volume of two spheres with different radii and the center identical to each molecule's center of gravity. We applied the descriptors to the classification of three types of chemotherapeutic drugs by using two pattern recognition methods, Soft Independent Modeling of Class Analogy (SIMCA) and the combination of genetic algorithms (GAs) and nonhierarchical clustering (NHC). The classification and prediction results that were independently compared through the two methods showed that the latter method (84.7% correctly classified and 52.0% correctly predicted) could give a better model than the former method (80.0% correctly classified and 48.3% correctly predicted).