Fig. 3From: MUMAL2: Improving sensitivity in shotgun proteomics using cost sensitive artificial neural networks and a threshold selector algorithmAnalysis of a ROC curve obtained from a model built with dataset S2_NPH_CH2. For each point of the curve a, the threshold and several statistical measures b related to that point are known. False and true positives are, respectively, instances of class 0 (decoys) and class 1 (targets) that were considered positives (P ≥ threshold) by the model. In c, it is shown how to estimate FDR among target hits for a given discriminant probability (threshold) of the model. It is simply the number of FPs over the number of TPsBack to article page