The .RAW data files obtained from the mass spectrometer were converted to .DTA files by use of extract_msn.exe, a windows console application provided by Thermo Electron (ThermoElectron Finnigan, San Jose, CA, USA). Tandem MS spectra with more than five product ions were extracted to .DTA files and then merged into .MGF files by use of a Perl script. Spectra were not grouped based on precursor mass. The data set in .MGF format is available at http://www.massmatrix.net/download/ . The data set was then searched by use of MassMatrix against the NCBInr human databases with the following options: i) Modifications: variable acetylation of lysine, variable acetylation of N-terminus; ii) Enzyme: trypsin; iii) Missed Cleavages: 3; iv) Peptide Length: 4 to 30 amino acid residues; v) Precursor Ion Charge: 1+, 2+, 3+; and VI) A mass tolerance of 0.02 Da and 0.01 Da for the precursor and product ions respectively. The same data set was also evaluated by Mascot, SEQUEST (SEQUST v.28 on BioWorks 3.3), X!Tandem, and OMSSA. The search parameters in Mascot, SEQUEST, X!Tandem, and OMSSA were identical to those in MassMatrix where appropriate. Critical values for scores in the three programs were set as follows: pp or pp2 value > 6 in MassMatrix; score > 30 in Mascot; XCorr > 1.5 for +1 peptides, 2.0 for 2+ peptides and 2.5 for 3+ peptides in SEQUEST; expectation value < 0.1 in X!Tandem; and e-value < 0.1 in OMSSA. If multiple peptide matches were found for a given spectrum only the match with the highest score was considered. In order to improve the performance of SEQUEST, the protein database was indexed prior to database searches.
The true positives and false positives for the three algorithms were determined by searches against a human database containing reversed decoy sequence as described by Elias [34 (link)]. The total number of false positive peptide matches was calculated by multiplying the number of peptide matches to reversed sequences by two. The number of true positives was then calculated by subtracting total number of false positives from the total number of peptide matches in the forward and reversed databases [34 (link)].
The true positives and false positives for the three algorithms were determined by searches against a human database containing reversed decoy sequence as described by Elias [34 (link)]. The total number of false positive peptide matches was calculated by multiplying the number of peptide matches to reversed sequences by two. The number of true positives was then calculated by subtracting total number of false positives from the total number of peptide matches in the forward and reversed databases [34 (link)].
