Development of Computational Methods for Cancer Research: Strategies for closing the feedback loop in omics workflows

As the ultimate workhorses of the living things, proteins undergo significant regulatory activity throughout the lifetime of a cell or an organism. Many complex diseases effect the protein composition, expression or modification in the cells or tissues they arise in. It is then no surprise that proteomics is a field full of promise, one which is expected to generate significant insights towards functional characterization of cancer and deliver potential targets of diagnostic, prognostic or therapeutic value. It is also a science in its teens, which still grows and keeps changing as it grows with the technological advances in instrumentation as the driving force. Since data analysis routines are yet to be established fully, functional characterization of protein expression regulation in cancer remains an open question. The work presented in this thesis provides an overview of the field based on technological, computational and biological aspects in the introductory chapters, introduces a novel method for functional evaluation of changes in protein expression and demonstrates its utility in PAPER I, and describes the insights gained from investigating proteomes of several different types of human malignancies. PAPER I underlines the challenges in functional analysis of expression data, especially from LC-MS/MS experiments, and describes a method based on a relatively simple mathematical model, which is used for subsequent analyses. PAPER II describes a study on soft-tissue sarcomas, with the proteomic analysis revealing insights to protein expression patterns potential differentiation paths. PAPER III demonstrates a pairwise comparison of malignancies of gastroesophageal track and corresponding normal tissue; we highlight several proteins and pathways as likely targets of expression regulation. PAPER IV and V on the other hand focus on breast cancer. PAPER IV presents results from an investigation of immortalized breast cancer cell lines and raises the question of how well these model systems represents the tumours they are expected to be alike. PAPER V demonstrates the therapeutic potential of inhibiting oestrogen signalling in ER+ breast cancer, using a luminal type patient-derived xenograft mouse model. Collectively, this thesis presents some of the key concepts in quantitative proteomics workflows, elaborates on the importance of data processing routines and through the papers in the appendix demonstrates the potential of functional analysis algorithms in generating insights to cancer biology