Broad assessment of robustness under different levels of data noise. Full range of result values of our algorithm in the presence of 0%,1% and 5% noise, after running triplicates of each one of the 60 sets of EA parameters. A) Minimum and maximum Recall values under different levels of noise. B) Minimum and maximum False Positive Rate (FPR) values. For our multi-parameter method, we observe that across all the ranges of values of results, the ratio between Recall/FPR on each one of the different levels of noise is always greater than 1, thus above the line of no discrimination within the Receiver Operating Characteristic (ROC) space.