In shotgun proteomics, false discovery rate (FDR) estimation is a necessary step to ensure the quality of accepted peptide-spectrum matches (PSMs) from a database search. Popular statistical validation tools for FDR control tend to rely on target-decoy searching to build empirical, dataset-specific models, which often leads to inaccurate FDR estimates. In this paper, we propose a new approach named common decoy distribution (CDD) to FDR estimation using the idea of a fixed empirical null score distribution derived from millions of peptide tandem mass spectra. To demonstrate the viability of CDD, its stability with respect to noise and the presence of unexpected peptide modifications was evaluated. PeptideProphet-based implementation of CDD was benchmarked against decoy-based PeptideProphet, and both methods exhibited similar accuracy of FDR estimates and retrieval of correct PSMs. The finding of this study calls for a re-evaluation of the necessity of dataset-specific target-decoy searches and illustrates the potential of Big Data approaches for statistical analysis in proteomics.
Keywords: false discovery rate; peptide identification; target-decoy search.