Semi-supervised learning for peptide identification from shotgun proteomics datasets

Abstract

Shotgun proteomics uses liquid chromatography-tandem mass spectrometry to identify proteins in complex biological samples. We describe an algorithm, called Percolator, for improving the rate of confident peptide identifications from a collection of tandem mass spectra. Percolator uses semi-supervised machine learning to discriminate between correct and decoy spectrum identifications, correctly assigning peptides to 17% more spectra from a tryptic Saccharomyces cerevisiae dataset, and up to 77% more spectra from non-tryptic digests, relative to a fully supervised approach.

Quantitative global phosphoproteomics of human umbilical vein endothelial cells after activation of the Rap signaling pathway

“…After identification and quantification, we processed all phosphoproteomic and proteomics results according to very stringent peptide acceptance criteria. These criteria include: (i) a false discovery rate (FDR) of less than 1% at the peptide level utilizing the Percolator-based algorithm, 33 (ii) peptides needed to have a Mascot ion score of at least 20 and (iii) peptide length was restricted between 6 and 45 residues. Subsequently, we defined two classes of phosphopeptides and phosphosites: a phosphopeptide was designated Class I when the PhosphoRS site probability for each phosphorylated residue within the phosphopeptide was at least 75%.…”

Extracellular vesicles from bone marrow‐derived mesenchymal stromal cells support ex vivo survival of human antibody secreting cells

“…We used the following parameters for database searching: 20 ppm precursor mass tolerance; fully digested with trypsin; up to three missed cleavages; fixed modification: carbamidomethylation of cysteine (+57.0214); variable modifications: oxidation of methionine (+15.9949). False discovery rates (FDRs) of peptide and protein identifications were evaluated and controlled to less than 1% by the target-decoy method [57] through linear discriminant analysis (LDA) [58]. Peptides fewer than seven amino acid residues in length were deleted.…”

Carcinogenesis of renal cell carcinoma reflected in HLA ligands: A novel approach for synergistic peptide vaccination design

“…Oxidized methionine was allowed as a dynamic modification. False discovery rates (FDR) were determined by the Percolator algorithm 60 based on processing against a decoy database consisting of the shuffled target database. FDR was set at a target value of q 0.05 (5% FDR).…”