Figure 5

Comparison of Caipirini with MedlineRanker in ROC space. Using the cell cycle data set as benchmark Caipirini can be compared to other tools - When compared with MedlineRanker, the two tools performed somewhat alike, although Caipirini seems to be slightly better for this dataset. ROC space: recall (sensitivity) versus false positive rate for the same dataset when the threshold is set at different scores. MCC*100: to describe also with a single measure the quality of the binary (i.e., two-class; A vs. B) classifications we used Matthew's correlation coefficient (MCC) - Generally, it is regarded as a balanced metric that can be used even if the classes are of very different sizes. MCC values range between -1 and +1 (Coefficients valued +1, 0, and -1 represent a perfect, an average random, and an inverse prediction, respectively); in the graphs the MCC value for each score has been multiplied by one hundred.