Spivak, M. et al. Mol. Cell. Proteomics advance online publication (3 November 2011).

In shotgun proteomics analyses, proteins are identified by comparing the experimental peptide mass spectra to theoretical peptide mass spectra generated from sequence databases. A crucial step in this approach is inferring the identity of the protein from these peptide-spectrum matches. The false discovery rate is typically controlled at the peptide but not protein level, which can result in error inflation. Spivak et al. describe a software tool, Barista, designed to identify proteins from shotgun mass spectrometry data with high confidence. Barista is based on a machine-learning algorithm that maximizes the number of proteins that can be confidently accepted as correct and minimizes the number of incorrect identifications. Barista outperformed current popular software tools including ProteinProphet and IDPicker, especially for identifying short proteins and proteins with only one detected peptide.