Figure 7. Mixture model approach for computing posterior probabilities.

All MS/MS spectra from the entire experiment are searched against a protein sequence database (without the need to append decoy sequences). The best database match for each spectrum is selected for further analysis. The most likely distributions among correct (dotted line) and incorrect (dashes) PSMs are fitted to the observed data (solid line). A posterior probability is computed for each peptide assignment in the dataset. The parameters of the distributions, including the mixture proportion π₁ are learned from the data using e.g. the EM algorithm.