Figure 7

Decision tree for WP vs . UP. Visualization of Well-predicted (WP) vs. Under-predicted (UP) decision tree, demonstrating the sequence determinants that affect the theoretical prediction at the N2 bond between the second and third amino acid positions. The first number in parenthesis following the classification (either UP or WP) is the number of true positive training points for that rule; that is, the number of instances that are covered by the rule. The subsequent number after the slash indicates the number of counter-instances to the rule; that is, the number of training points that are covered by the rule but are not the class indicated. The four paths through the tree illustrated via bolded black edges and non-filled boxes outlined in red are the sequence determinants leading to under- and well-prediction in the theoretical model. The numbers of data points in the tree do not sum to the total shown at the top due to node pruning.