We have investigated if phosphopeptide identification and simultaneous site localization can be achieved by spectral library searching. This allows taking advantage of comparison of specific spectral features, which would lead to improved discrimination of differential localizations. For building a library, we propose a spectral simulation strategy where all possible single phosphorylations can be simply and accurately (re)constructed on enzymatically dephosphorylated peptides, by predicting the diagnostic fragmentation events produced in beam-type CID. To demonstrate the performance of our approach, enriched HeLa phosphopeptides were dephosphorylated with alkaline phosphatase and analyzed with higher energy collisional dissociation (HCD), which were then used for creating a spectral library of simulated phosphopeptides. Spectral library searching using SpectraST was performed on data sets of synthetic phosphopeptides and the HeLa phosphopeptides, and subsequently compared to Mascot and Sequest database searching followed by phosphoRS and Ascore afforded localization, respectively. Our approach successfully led to accurate localization, and it outperformed other methods, when phosphopeptides were covered by the library. These results suggest that the searching with simulated spectral libraries serves as a crucial approach for both supplementing and validating the phosphorylation sites obtained by database searching and localization tools. For future development, simulation of multiply phosphorylated peptides remains to be implemented.
Keywords: LC−MS/MS; beam-type CID; phosphoproteomics; phosphorylation site localization; spectral library searching; spectral simulation.