Database normalization is crucial for reliable protein identification in mass spectrometry-based proteomics

Abstract

Research in proteomics is driven by mass spectrometry, especially the identification of proteins from complex samples. Computational analysis of the resulting data determines the peptide sequences of the recorded spectra and integrates identifications into proteins. For this, database search algorithms can be employed, but they need a list of amino acid sequences that are expected to exist in the sample. Many algorithms have been proposed and consensus scoring has been performed. While the comparison/integration among results from different algorithms is important, there has been no attempt to integrate the results from searching multiple databases. This is, however, important since it poses technical problems when all databases, needed for a study, are simply concatenated. Unfortunately, it has been shown that databases of different size influence scoring and prohibit the direct comparison of results.