Feature detection using high-resolution mass spectrometry: Software and applications for shotgun proteomics.

摘要

Technological advances in the field of mass spectrometry have produced hybrid instruments that combine high-resolution Fourier transform mass spectrometry with the speed of linear ion trap mass analyzers. When used for shotgun proteomics, these hybrid instruments supplement MS/MS data acquisition with high-resolution and high mass accuracy precursor spectra; however, existing software is poorly suited for effective analysis of these high-resolution data in a shotgun proteomics context. The goal of this research is to design a fast and robust algorithm for the analysis of high-resolution mass spectra and demonstrate its relevance in the use of shotgun proteomics.;The software algorithm, Hardklor, simplifies high-resolution mass spectra to identify the charge states and monoisotopic masses of observed peptide signals. In an analysis of Escherichia coli, Hardklor detected 244,146 total peptide isotope distributions over a 75 minute gradient on an LTQ-Orbitrap mass spectrometer. These data represent 15,309 distinct, chromatographically persistent peptide isotope distributions. Hardklor also detected peptide isotope distributions resulting from the labeling of peptides with 18O heavy isotopes.;The ability of Hardklor to detect peptide signals from precursor spectra was used to improve the performance of shotgun proteomics. Hardklor-detected precursor signals that were missed by data-dependent acquisition were targeted in an automated, iterative methodology known as post analysis data acquisition (PAnDA). PAnDA produced a 30.9% increase in peptide identifications over standard data-dependent acquisition of a Caenorhabidits elegans sample. The accurate mass values determined by Hardklor were also used to improve database searching performance. An analysis of tandem mass spectra from Saccharomyces cerevisiae produced 127% more peptide-spectrum matches using an accurate mass search at 5 ppm instead of a standard 3 Da wide mass tolerance window. Hardklor algorithm features were used to provide information about protein structure information using chemical crosslinking and 18O-labeling. Structural information was obtained for beta-lactoglobulin, GFP, lysozyme, RNase A, and the two interacting domains of the BRCA1/BARD1 RING-domain heterodimer. Also, the Hardklor algorithm was used to identify sites of N-linked glycosylation from membrane protein fractions of Saccharomyces cerevisiae. Five different strains were profiled for N-linked glycosylation and 72 glycosylated peptides showed significant differences in expression levels among the strains.