Statistical methods for mass spectrometry-based clinical proteomics

The work presented in this thesis focuses on methods for the construction of diagnostic rules based on clinical mass spectrometry proteomic data. Mass spectrometry has become one of the key technologies for jointly measuring the expression of thousands of proteins in biological samples. However, the development of MS instrumentation gave rise to new statistical challenges in the processing and analysis of the acquired data. This is due to the complex nature of the spectral proteomic signal which is measured, as it consists of high-dimensional functions representing the within-patient proteome expression. This work considers new methods to respond to these challenges. Our main interest focuses on the comparison of mass spectral proteomic profiles collected from healthy individuals and cancer patients in the context of distinct case-control studies. A key objective in such studies is the construction of discriminant rules for distinguishing between individuals as to the presence or absence of the disease as well as for predicting the health status of future patients. We present a series of data analyses for distinct case-control cancer studies where we address these questions through the use of methodology specific for the type of data considered in each of the studies.