Figure 2.

Two-step database search in combination with target-decoy strategy leads to a dramatic increase in false positive rate. (a) The protein groups count is shown for single- or two-step search strategies across increasingly large protein sequence databases. Counts are color-coded per category, with eukaryote (gray), bacteria (red), contaminant (blue) and reverse (Orange) hits. (b) The FDR is calculated for single- or two-step search strategies across increasingly large protein sequence databases. The FDR is calculated based on reverse hits only (circle shape) or reverse plus bacterial hits (triangle shape). (c & d) The sensitivity (c) and factual FDR (d) based on protein groups identification across increasingly large protein sequence databases. The compared database search strategies are single-step (blue), two-step taxon filtering (gray) and two-step protein filtering without (red) or with (Orange) database sectioning. Lines represent the median (and the shading corresponds to the standard error) from N = 8 LC-MS/MS runs. (e) The true positive count based on protein groups identified with a minimum of one (shaded coloring) or two (unshaded coloring) unique peptides for the largest database (i.e. 20). The compared database search strategies are single-step (blue), two-step taxon filtering (gray) and two-step protein filtering without (red) or with (Orange) database sectioning. Bars and numbers indicate the median count, while error bars correspond to the standard deviation, from N = 8 LC-MS/MS runs. The overall maxima of true positive count based on single-step search is indicated as a horizontal dotted line (gold)