Combining percolator with X!Tandem for accurate and sensitive peptide identification

In this work, Percolator was successfully interfaced with X!Tandem using a PHP program to generate an improved search platform, X!Tandem Percolator. In order to achieve the best classification performance of peptide identifications in Percolator, a set of experimentally validated spectral identifications (34,993 MS/MS spectra) were used to guide the development of discriminatory features from X!Tandem search results. By comparing the features (e.g., Log(E) and mass error) of these experimentally validated peptide matches with those of false identifications, a comprehensive set of features can be chosen for Percolator in an objective and rational manner. The accuracy of X!Tandem Percolator was demonstrated by comparing the estimated q-value of the validated data set with the empirical q-value. By comparing the results from the X!Tandem Percolator and the original X!Tandem, superior sensitivity and specificity of the X!Tandem Percolator result was demonstrated on various shotgun proteomic data sets under different search conditions. In all of the cases studied in this work, X!Tandem Percolator could improve the number of peptide identifications at the same level of q-values.