Classification of Aurora kinase inhibitors by self-organizing map (SOM) and support vector machine (SVM)

Abstract

The Aurora kinase family (consisting of Aurora-A, -B and -C) is an important group of enzymes that controls several aspects of cell division in mammalian cells. In this study, 512 compounds of Aurora-A and -B inhibitors were collected. They were classified into three classes: dual Aurora-A and Aurora-B inhibitors, selective inhibitors of Aurora-A and selective inhibitors of Aurora-B by Self-Organizing Map (SOM) and Support Vector Machine (SVM). The prediction accuracies of the models (based on the training/test set splitting using SOM method) for the test set were 92.2% for SOM1 and 93.8% for SVM1, respectively. In addition, the extended connectivity fingerprints (ECFP_4) for all the molecules were calculated and structure-activity relationship of Aurora kinase inhibitors was summarized, which may be helpful to find the important structural features of inhibitors relating to the selectivity to Aurora kinases.