Figure 5.

A calibrated score, on average, yields more discoveries (Käll). Each panel plots the median number of discoveries as a function of FDR threshold using the Käll method applied to the raw and calibrated scores (the median is with respect 1000 applications, each using a single independently drawn decoy set). For small FDR values, the calibrated score yields considerably more discoveries than the uncalibrated score. Note that even for MS-GF+ E-value our calibration procedure typically increases the number of Käll discoveries.