A systematic statistical analysis of ion trap tandem mass spectra in view of peptide scoring

Abstract

Tandem mass spectrometry has become central in proteomics projects. In particular, it is of prime importance to design sensitive and selective score functions to reliably identify peptides in databases. By using a huge collection of 140000+ peptide MS/MS spectra, we systematically study the importance of many characteristics of a match (peptide sequence/spectrum) to include in a score function. Besides classical match characteristics, we investigate the value of new characteristics such as amino acid dependence and consecutive fragment matches. We finally select a combination of promising characteristics and show that the corresponding score function achieves very low false positive rates while being very sensitive, thereby enabling highly automated peptide identification in large proteomics projects. We compare our results to widely used protein identification systems and show a significant reduction in false positives. © Springer-Verlag Berlin Heidelberg 2003.