In a publication by third party researchers, PEAKS reliably outperformed other commercial and open-source methods for peptide identification. Across all four samples of HLA class 1 peptides, PEAKS located the greatest number of identifications, even under strict 1% pep-level FDR (corrected to 0.5%), Figure A.
Figure B presents a five-set venn diagram which further demonstrates the significant coverage and performance. PEAKS is shown to identify significantly more (195) correct peptides, compared to 83 by the next leading scorer (Mascot + Percolator) when set to 5% FDR (corrected to 2.5%).
A researcher not connected to the authors further supported the PEAKS findings by stating “These numbers are pretty consistent with what we find with HLA peptides which are non-tryptic.” Curtis McMurtrey, University of Oklahoma Health Sciences Center.
The publication can be found here: http://elifesciences.org/content/4/e07661.