The 100-fold-CV performance of the 478 classifiers that are searched during the second classifier selection.

Highlighted in yellow are the three classifiers that satisfied the criterion. Columns under “Leave-one-exp-out” and “Four-exp-combined” correspond to the performance in the leave-one-experiment-out setting and four-experiments-combined setting, respectively. In the leave-one-experiment-out setting, a different experiment is used as the validation set and the columns D-K denote the specificity and sensitivity values on the four validation sets. Columns L and M denote the average of the specificity and sensitivity values in columns D-K, respectively. Columns N and O denote the minimum of the specificity and sensitivity values in columns D-K, respectively. In the four-experiments-combined setting, four experiments are combined into a dataset. Columns P-W denote the specificity and sensitivity values calculated using only the samples from a single experiment. Columns X and Y denote the average of the specificity and sensitivity values in columns P-W, respectively. Columns Z and AA denote the minimum of the specificity and sensitivity values in columns P-W, respectively. Columns AB and AC denote the sensitivity and specificity values computed in the usual way i.e., using all the samples. (XLSX)