Fig. 6

FDR\(_{prot}\) computed from \(\hat{s}\) values for proteins with at least 6 mapped PSMs in the SIHUMIx data set (see Methods). Two strategies are used to compute the FDR\(_{prot}\), the “classical” and the “picked” strategy (see [29] and the “Methods” section). The regression lines extrapolating to large value of \(\hat{s}\), for which the decoy database produced no hits, is computed from the \(\hat{s}\)-score interval [1.0, 2.5] for SIHUMIx and [0.75, 2] for \(0.75-2\) E. coli. The cutoff-value \(\hat{s}=3.5\) for SIHUMIx corresponds to a q-value of 0.03. For the much smaller E. coli data set, the same q-value is is obtained for \(\hat{s}\approx 2.7\)