Decoy Generation Method for Confident Peptide Identification in Spectral Library Search

Abstract

As an alternative to sequence database search for identifying peptides from tandem mass spectrometry (MS/MS), spectral library search is gaining momentum owing to its high sensitivity and fast speed. In sequence database search, a query spectrum is compared with hypothetical spectra generated from the database of peptide sequences. On the other hand, spectral library search matches a query spectrum with the spectra that have been already identified in previous analyses. Spectral library search is more sensitive than database search because it utilizes peptide fragment patterns. In spectral library search, false positives can be estimated using the target-decoy method, where decoy spectra should not be biologically present but should share the characteristics of target spectra. Various decoy generation methods have been proposed to accurately estimate the number of false positives. In this study, we present a new decoy generation method for accurate estimation of false discovery rate (FDR) in spectral library searches. For various target and decoy libraries, their target and decoy biases were evaluated using null query spectra and the resulting FDRs were compared. We demonstrated that our spectral library search model could remove the target bias and could accurately estimate the number of false positives.