You could try new approaches using the models we published here, or compare them with other classifiers (matlab, python, anything you want!). Also, new testing data would be welcome to evaluate the models.