Figure 5: Improved version of average target-decoy competition reduces conservative bias for low FDR thresholds.

(A) The figure plots, for the Kim dataset applied to a database of 1000 proteins, the number of PSMs accepted as a function of FDR threshold for 10 runs of TDC⁺ and 10 runs of aTDC⁺. (B) Same as panel A, but using the improved aTDC⁺ procedure (called “${\text{aTDC}}_1^{+}$” in the main text). In both panels, each run of aTDC⁺ or ${\text{aTDC}}_1^{+}$ is computed with respect to 10 decoy databases.