T. Rögnvaldsson, J. Häkkinen, C. Lindberg, G. Marko-Varga, F. Potthast, and J. Samuelsson
Improving Automatic Peptide Mass Fingerprint Protein Identification by Combining Multiple Peak Sets
Submitted to Analytical Chemistry

Abstract
Several automatic peak detection strategies are presented and compared against manual peak picking, on a set of mass spectra obtained after tryptic in gel digestion of 2D-gel samples from human fetal fibroblasts. The spectra represent samples with both high and low protein abundance, and it is shown that peak sets selected by human operators have a considerable variability and that it is impossible to speak of a single true peak set. A completely automated peak detection strategy is developed that avoids time-consuming parameter tuning, and which yields protein identification efficiency that exceeds the human operator. The described method therefore allows automated user-independent protein identification using peptide mass fingerprints.

LU TP 03-04