Computational mass spectrometry : algorithms for identification of peptides not present in protein databases

Mass spectrometry has revolutionized protein identification in the last decade. Efficient algorithms have been developed to identify peptides that are encoded in protein databases. This work presents novel methods for interpretation of mass spectrometry data on compounds that are not directly encoded in protein databases. One example of compounds that are not in protein databases are nonribosomal peptides. Because of the specialized machinery that synthesizes these compounds and their unique (often cyclic) structure, traditional database search tools cannot analyze these data. With new algorithmic developments, we show that mass spectrometry can speed up the process of characterization of cyclic peptides which are compounds of great interest in drug discovery. A second class of peptides that are not encoded in protein databases are peptides that are the product of fused proteins. This type of peptides can arise from cancer proteomes, where the peptide spans a fusion point. Again, traditional search tools cannot identify fusion peptides because they are not directly encoded in the databases. We also present an algorithm to identify such peptides. Finally, mutated and modified peptides are also not directly encoded in protein databases. Existing tools are particularly inefficient when searching for unexpected modifications. Although this problem has been addressed with "blind" database search tools, their running time is too demanding to be practical for large databases. We develop new methods to search for mutated and modified peptides that are orders of magnitude faster than existing tools. Overall, with new algorithmic developments we enable mass spectrometry to characterize novel compounds that evade identification with traditional MS/MS tool