Figure 3

Protein inference results of Mix 1. This figure shows the protein inference results of Mix 1. The protein identification threshold is set as 0. 95. Figure 3 (a) gives the number of true positive proteins identified by the iterative method (IM) and ProteinProphet. It can be seen that IM always identifies more true proteins than ProteinProphet. When the high-confidence peptide selection threshold is less than 1, all the 31 true proteins can be identified; while 30 proteins can be identified when the threshold is 1. This indicates that all peptides identified for the missed protein have probabilities less than 1. Meanwhile, ProteinProphet can only identify 27 of the 31 true proteins. Figure 3 (b) shows that the number of false positive proteins output by IM decreases with the increase of high-confidence peptide selection threshold, however, ProteinProphet outputs the same number of false positive proteins regardless of the threshold. The reason is that higher selection thresholds reject more false peptides to be included in the iteration process, and thus less false proteins will be identified by IM. On the other hand, some false positive proteins are removed with the elimination-rule in the iteration process, while they are retained by ProteinProphet.