Figure 1
From: Feature-based multiple models improve classification of mutation-induced stability changes
Design of feature-based multiple models. The design of EASE-SS is shown, however, the same applies to EASE-ASA. First, the data was divided according to the predicted secondary structure (accessible surface area). Then, relevant predictive features were selected using a greedy feature selection algorithm. SVM parameters were optimised using a grid search. Finally, the predictive models were trained.