Optimized data analysis avoiding trypsin artefacts

Most bottom-up proteomics experiments share two features: The use of trypsin to digest proteins for mass spectrometry and the statistic driven matching of the measured peptide fragment spectra against protein database derived in silico generated spectra. While this extremely powerful approach in combination with latest generation mass spectrometers facilitates very deep proteome coverage, the assumptions made have to be met to generate meaningful results. One of these assumptions is that the measured spectra indeed have a match in the search space, since the search engine will always report the best match. However, one of the most abundant proteins in the sample, the protease, is often not represented in the employed database. It is therefore widely accepted in the community to include the protease and other common contaminants in the database to avoid false positive matches. Although this approach accounts for unmodified trypsin peptides, the most widely employed trypsin preparations are chemically modified to prevent autolysis and premature activity loss of the protease. In this study we observed numerous spectra of modified trypsin derived peptides in samples from our laboratory as well as in datasets downloaded from public repositories. In many cases the spectra were assigned to other proteins, often with good statistical significance. We therefore designed a new database search strategy employing an artificial amino acid which accounts for these peptides with a minimal increase in search space and the concomitant loss of statistical significance. Moreover, this approach can be easily implemented into existing workflows for many widely used search engines.