Figure 7

Peptide identification results of Mix 1. This figure illustrates the peptide identification results of Mix 1. Figure 5 (a) shows that IM outputs all true peptides at any threshold but 1, while the number of true peptides output by PeptideProphet deceases with the increase of threshold. The reason is that IM recomputes the probability of false negative peptides by using extra information from the identified proteins. The fact that it cannot output all true peptides at the threshold of 1 agrees with that one true protein is missed at this threshold. Figure 5 (b) shows that IM can produce much less false positives than PeptideProphet at all thresholds but 1. At threshold 1, although IM outputs 8 false positive proteins (see Figure 3(b)), less than 16 from ProteinProphet, its generation of more false positive peptides is due to the feedback framework which unifies the peptide and protein identification results. That is, negative peptides from false positive proteins are output as false positives, while for PeptideProphet, peptides are selected only by their probabilities. The feedback framework also explains why the number of false positives from IM is steady versus threshold, that is because these peptides are from the few counterpart proteins.