Benchmarking Glycoproteomics Software Using Diagnostic Ion Based Post-Validation for Glycopeptide Identification

Advancements in glycoproteomics software have improved glycopeptide identification. However, algorithmic differences cause discrepancies in glycopeptide identifications even when identical data sets are used. We compared five state-of-the-art glycoproteomics software programs (Byonic, MSFragger-Glyco, pGlyco3, Glyco-Decipher, and GRable), investigating their ability to identify rare NeuGc- and KDN-containing glycopeptides from BJAB-K20 cells, which lack UDP-N-acetylglucosamine 2-epimerase, the rate-limiting enzyme for sialic acid synthesis. Approximately one-third of the identified glycopeptides were unique to individual tools. Byonic identified the most glycopeptides, whereas Glyco-Decipher and GRable identified complex, highly branched glycans. NeuGc- and KDN-containing glycopeptides were identified by specific programs, highlighting their ability to handle rare glycan structures. To assess the reliability of these identifications, we reanalyzed the MS/MS spectra and stratified all identifications retained under the software-specific criteria based on spectral evidence from diagnostic oxonium ions and Y0/Y1 or b/y ions. Although most were robustly supported, some relied on weaker evidence and warranted cautious interpretation. Leveraging the capabilities of each program enabled comprehensive, reliable analysis of glycopeptides with rare glycan structures. Combining software programs with complementary strengths and systematic postvalidation using diagnostic oxonium ions and Y0/Y1 ions provides an independent spectrum-level confidence assessment beyond conventional FDR-based filtering, improving the reliability and depth of glycopeptide identification.