Protein Level Probabilitiesfor Shotgun Proteomics Experiments

Sammanfattning: There exist many techniques to perform the analysis of proteins. Current Shotgun Proteomics based methods assign peptide level scores to peptide-spectrum matches obtained by matching observed spectra against a database of theoretically generated spectra from a set of known proteins. This set of peptide-spectrum matches is ranked according to a score based on the quality of the match. Subsequently, the candidate present proteins can be inferred from the confidently identified peptides. This process seems straightforward and out of the box, however, it has some notable weaknesses. The imperfections of the set of scores given by the database search engine tool lead to the need to apply a post-processing tool that can give more accurate scores yielding more precise information of the number of peptides believed to be present in the sample. However, the set of proteins believed to be present in the analyzed sample should not be inferred directly from the set of high scored peptides. For example, in cases where there are peptides that form part of more than one protein, or when there is a protein that contain both high and low scored peptides, or when there is a high scored peptide that indicates that the protein is present but it is a false positive, etc. Therefore, there is a need for a tool to infer the list of confident proteins from the set of scored peptides and assign confidence scores to the inferred proteins accounting for the problems described. We present in this thesis work a software tool for the analysis of proteins. This tool is based on the integration of the protein inference tool Fido with the postprocessor Percolator as a combined PSM post-processing and protein inference package that efficiently estimates protein level probabilities and confidence scores. We show that our integration of Fido and Percolator surpasses the stateof- the-art protein inference tool Protein prophet in terms of calibration, sensitivity, specificity and number of proteins correctly identified.