Figure 14. Computing protein probabilities (scores).

Posterior probabilities (scores) of peptides mapping to the same protein are utilized to compute the protein score. This score is then used to filter the list of protein identifications. Common approaches include taking the score of the highest scoring peptide, applying a multiple- (e.g. two-) peptide rule, computing p-values using a Poisson distribution-based model, or using combined evidence approaches. ProteinProphet utilizes the combined evidence approach, but with an additional adjustment of the initial peptide probabilities to account for non-random grouping of peptides to proteins (adjustment for the number of sibling peptides, NSP). In the case of multiple PSMs identifying the same peptide sequence, typically only the score of the best PSM is used at the protein level. As an option, the unmodified and a modified version of the same peptide can be treated as different peptides, with both contributing to the protein score.