Detailed classification performances per dataset, SPARTA run, and selection iteration.

The first sheet contains the detailed information as plotted in Fig 3: the average AUC scores, per run, for the overall best iteration level, for each dataset (taxa and annotations). The median of the average values, and p-values of the Mann-Whitney U-test comparisons between the taxon and FA average scores per disease are also given. P-values under the 0.05 threshold are considered significant and are highlighted with a *. Other sheets contain the details of each RF trained per run and iteration for each dataset (read: [dataset]_R_[run number]_It_[iteration number], with iteration numbers initialized at 0). The information given per sheet is: for each of the 20 RFs trained in this iteration and run, the optimal parameters found through GridSearch, the optimal threshold for probability prediction, and the AUCs on the training, validation, and test subsets.
(ZIP)