Exploiting Proteomics Data

Abstract

Proteomics plays a central role in understanding complex disease mechanisms, especially since it is well known that the effectors of biological functions are mostly proteins. Beside classical gel-based techniques especially Mass Spectrometry (MS) has emerged as the standard technique for proteomics experiments. MS-based proteomics has evolved into several different and partly complementary technologies. In this thesis we have analyzed data generated by the three complementary technologies: Matrix-Assisted Laser Desorption/Ionization (MALDI), Isobaric Tags for Relative and Absolute Quantitation (iTRAQ) and 2D Difference Gel Electrophoresis (DIGE). The three technologies are applied to an obesity-induced mouse model in order to gain relevant knowledge on biological processes involved in diabetes. The primary goal of this thesis is to develop and implement specifically tailored data analysis methods for each technology with the aim to improve quality and reliability of the results compared to standard evaluation workflows. The developed methods benefit from the fact that in proteomics a single protein is typically represented by several peptides showing more or less similar measurements. Combining this similarity information and advanced statistical testing, we are able to identify sets of potential biomarkers that may play an important role in diabetes disease mechanisms. The identified biomarkers are very well suited for building a classification engine to predict disease relations. However, peptides derived from the same protein may also show contradictory quantitations (e.g. a protein is two-fold up regulated and two- fold down regulated at the same time). This could be due to technical artifacts or biological properties (e.g. protein isoforms). We try to resolve these contradictions with PPINGUIN, a workflow developed for the reliable quantitation of iTRAQ experiments. Application of the developed methods leads to improved results compared to standard data evaluation methods. The three technologies have a complementary character and therefore a direct comparison is difficult and shows only a small overlap. But a comparison based on the more abstract level of biochemical pathways shows a surprisingly good agreement of the results. In order to better understand the complex processes involved in diabetes a major challenge remains in integrating the results with other ’omics’ technologies.