Computational prediction and perturbation analysis of ceRNA networks in cancer
Post-transcriptional regulation of gene expression can be modeled as a competitive endogenous RNA (ceRNA) network, in which endogenous RNA transcripts (mRNAs, lncRNAs and circRNAs) compete for microRNA (miR) binding. Research shows that this competition maintains and fine-tunes levels of protein coding genes and the disruption of the network contributes to phenotypic conditions like disease (e.g. cancer). The objectives of this dissertation study are to 1) develop an algorithm and a user friendly online tool for construction of ceRNA networks for a gene of interest of a specific phenotype, 2) identify the main components of the network using network perturbation, and 3) develop algorithms for construction and perturbation analysis of genome-wide ceRNA networks and apply the proposed algorithms in breast cancer. The approaches to the study of the ceRNA phenomenon exploit concepts of miR regulation of gene expression, gene regulation modulation, and biological networks, which are the building blocks of our algorithms for construction and analysis of ceRNA networks. In Chapter Two, we present an online tool, TraceRNA, which is developed for construction of a ceRNA network of a gene of interest. The use of algorithms around individual genes as starting point in biomedical research has been an approach traditionally followed by investigators in the study of genomic functionality and its application to medicine. The outcome of the proposed approach is a list of the main components (genes and miRs) for the network based on that specific gene under study. In Chapter Three, we present an algorithm that applies network perturbation to search for the main components of the constructed ceRNA network (NetceRNA) of a specific phenotype. Perturbation in this study consists of the simulated change of expression of each one of the entities used to construct the original network (miRs, lncRNAs or mRNAs) and a measure of the differences between the original (e.g. normal) and perturbed network (e.g. disease). The components that perturb the network the most are considered as potential marker components of the specific phenotype. In Chapter Four, we present a mathematical model for construction, perturbation, and stability analysis of genome-wide ceRNA networks and apply the model to study breast cancer. We demonstrate in this dissertation models and algorithms for constructing and analyzing ceRNA networks. Our results in glioblastoma multiforme and breast cancer showed network components that can be considered as potential markers of the phenotype, which suggest that ceRNA networks could provide an alternative to the genomics markers for disease treatment.