Unbiased statistical analysis for multi-stage proteomic search strategies.

"Multi-stage" search strategies have become widely accepted for peptide identification and are implemented in a number of available software packages. We describe limitations of these strategies for validation and decoy-based statistical analyses and demonstrate these limitations using a set of control sample spectra. We propose a solution that corrects the statistical deficiencies and describe its implementation using the open-source software X!Tandem.