Inferring Proteolytic Processes from Mass Spectrometry Time Series Data

Proteolysis, the catalyzed hydrolysis of peptide bonds, is an important post-translational modi ication, having a signi icant in luence on the life cycle of protein and peptides. It is involved in numerous biological processes, like apoptosis, cell cycle progression, or blood coagulation. More then 500 genes were annotated as proteases, the enzymes catalyzing proteolytic cleavage of proteins and peptides, but many of them are still insuf iciently characterized. Hence a profound understanding of proteolytic processes is essential for a detailed analysis of many biological processes. Furthermore proteolysis is associated with multiple complex diseases like cancer and Alzheimer’s disease and is known to be involved in the infection with the HI-virus. Beyond its implication in biological processes, proteolysis can also be utilized for diagnostic and treatment purposes. Proteases, the enzymes catalyzing proteolytic cleavage, are established drug targets and their potential as biomarkers has been postulated in 2006 by Villanueva et al. In this thesis we present a novel approach to the characterization of proteolytic processes using mass spectrometry data. We utilize the qualitative and quantitative information of the mass spectra to construct a model, the degradation graph, containing all involved peptides aswell as the individual proteolytic reactions that connect them. We further propose a transformation of the degradation graph into a mathematical model that can be utilized in combination with the mass spectrometry data to estimate the rate constants of the individual reactions inside the degradation graph. Additionally we developed a score that can be used to rate different degradation graphswith respect to their ability to explain the observedmass spectrometry data. We use this score to iteratively improve the structure of an initially constructed degradation graph so as to account for errors during the construction of the degradation graph. Whilemore andmoremass spectrometrydata is producedand is publicly available, there is a lack of well annotated, so called gold standard or ground truth datasets. Those datasets are required for a thorough benchmarking of novel algorithms and newly developed software. This problem is increasing as the experimental setups and scienti ic questions in computational mass spectrometry get more and more complex. We thereforepresentMSSimulator, a comprehensive simulator formass spectrometrydata. Although using simulated data does not remove the need for testing on real datasets, it eases algorithmbenchmarking and development, due to the availability of ground truth datawhich enables us to compare and validate the results more effectively. MSSimulator is the currently