Figure 2

External validation of geNetClassifier using a large dataset of leukemia samples. (A) Assignment of 200 leukemia samples in one run of external validation. Green dots correspond to samples in which the most likely class is correct and red dots to samples in which it is incorrect. Samples in the blue background area are assigned to the most likely class; samples under any of the two thresholds (dashed blue lines) are considered doubtful samples and can be left as "Not-Assigned". (B) Summary of the results of 10 independent external validations. Average confusion matrices (in percentage values) and statistical parameters: sensitivity, specificity, MCC, call rate (calculated per class), global accuracy and global call rate (calculated globally for all the classes). On the right (labelled All Assigned ) the statistics are calculated assigning all the samples to their most likely class. On the left (With Not-Assigned ) the statistics are calculated leaving doubtful samples unassigned.