Computational Biology in Clinical Proteomics and Chromatin Genomics

Abstract

The work in this thesis is concerned with two very distinct biological fields. The first part pertains to the development of techniques to aid in the search for clinical biomarkers for use in the early detection of cancer. The second part aims to elucidate in what way a genome is organised in a cell nucleus and the functional consequences of this organisation. Part I: Clinical Proteomics. Cancer is a leading cause of death world-wide. The success of treatment is directly correlated with the stage of tumour progression. Therefore, it is of great importance to detect the occurrence of cancer as early as possible. Already for a long time, it is believed that the presence of a tumour has consequences for the repertoire of proteins and fragments thereof, i.e. peptides, present in the blood circulation. It has been proposed to use mass spectrometry to analyse the proteomic content of blood samples. Ultimately, such an approach would be used in routine population screening efforts, with the great advantage that the technique is largely non-invasive, as opposed to taking biopsies. After analysing samples using mass spectrometers, computational methods can be used to identify which peptides are predictive for a certain disease status. Such peptides are referred to as biomarkers. In Part I of this thesis, we describe work concerning the development of computational methods for processing mass spectrometry data with the goal of identifying such biomarkers. The first step in a mass spectrometry data analysis project is commonly the normalisation of data. Typically, raw same-sample mass spectra are not very comparable, due to high levels of inter-spectra variation. For this reason, spectra are normalised in an attempt to reduce this variance. We have conducted a comprehensive comparison of various normalisation methods, which are described in this thesis. We demonstrate that the method used by the majority of users performs very poorly, and advise on several methods that improve the performance significantly. After normalisation, spectral peaks representing the presence of peptides can be identified. In this thesis, we propose a method for doing so using multiple intermediate measurements that are normally discarded. We show that this approach outperforms existing methods and allows one to attach significance levels to detected peaks. Part II: Chromatin Genomics. All organisms are made up of cells; each cell containing an exact copy of the genome (i.e. the full collection of DNA). A large subgroup of organisms has their DNA contained in a separate compartment within the cell, called the nucleus. This subgroup of organisms, including animals like ourselves, is collectively referred to as eukaryotes. The diameter of a single human cell nucleus is about 6 micrometre, while the total length of all DNA contained in it is approximately 2 metres. This poses two interesting main questions. The first one is concerned with how this large amount of DNA is stored in such a confined space. Indeed, the three-dimensional organisation of chromosomes within the nucleus is largely unknown. The nucleus has a membrane separating it from the rest of the cell. The inside of this membrane is lined with a network of proteins collectively referred to as the nuclear lamina. In this thesis we present high-resolution maps of the interaction of human and mouse genomes with this nuclear lamina. We find that mammalian chromosomes are organised by way of large Lamina Associated Domains (LADs). In this way, we provide a detailed view of the spatial organisation of interphase chromosomes. The second main question is to do with the consequences of this organisation for the function of a cell. We find that during cell differentiation chromosomes are substantially refolded. In fact, hundreds of genes either migrate away from or towards the nuclear lamina during this process. We show that these genes change their activity upon relocalisation; genes that move towards the nuclear lamina are turned off, while genes that are removed from the lamina become more active. Despite this, we find that most of the spatial chromosome organisation is identical across all cell types we studied. We propose that these static regions collectively form a basal chromosome architecture and find that it is extremely well conserved between mouse and human, even though these species are separated in evolution by more than 75 million years. Using sequence analyses, we demonstrate that the basal chromosome architecture is largely encoded in the underlying genomic sequence. We further provide evidence that this genomic sequence alone is enough to tether specific regions to the nuclear lamina. Taken together, we show that mammalian genomes are organised in the cell nucleus by large regions contacting the nuclear lamina, which are largely static across cell types as well as between species. We further provide a potential mechanistic explanation in which the association of loci with the nuclear lamina is directly encoded in the genomic sequence.