Exploration of large molecular datasets using global gene networks : computational methods and tools

Abstract: Defining gene expression profiles and mapping complex interactions between molecular regulators and proteins is a key for understanding biological processes and the functional properties of cells, which is therefore, the focus on numerous experimental studies. Small-scale biochemical analyses deliver high-quality data, but lack coverage, whereas high throughput sequencing reveals thousands of interactions which can be error-prone and require proper computational methods to discover true relations. Furthermore, all these approaches usually focus on one type of interaction at a time. This makes experimental mapping of the genome-wide network a cost and time-intensive procedure. In the first part of the thesis, I present the developed network analysis tools for exploring large- scale datasets in the context of a global network of functional coupling. Paper I introduces NEArender, a method for performing pathway analysis and determines the relations between gene sets using a global network. Traditionally, pathway analysis did not consider network relations, thereby covering a minor part of the whole picture. Placing the gene sets in the context of a network provides additional information for pathway analysis, which reveals a more comprehensive picture. Paper II presents EviNet, a user-friendly web interface for using NEArender algorithm. The user can either input gene lists or manage and integrate highly complex experimental designs via the interactive Venn diagram-based interface. The web resource provides access to biological networks and pathways from multiple public or users’ own resources. The analysis typically takes seconds or minutes, and the results are presented in a graphic and tabular format. Paper III describes NEAmarker, a method to predict anti-cancer drug targets from enrichment scores calculated by NEArender, thus presenting a practical usage of network enrichment tool. The method can integrate data from multiple omics platforms to model drug sensitivity with enrichment variables. In parallel, alternative methods for pathway enrichment analysis were benchmarked in the paper. The second part of the thesis is focused on identifying spatial and temporal mechanisms that govern the formation of neural cell diversity in the developing brain. High-throughput platforms for RNA- and ChIP-sequencing were applied to provide data for studying the underlying biological hypothesis at the genome-wide scale. In Paper IV, I defined the role of the transcription factor Foxa2 during the specification and differentiation of floor plate cells of the ventral neural tube. By RNA-seq analyses of Foxa2-/- cells, a large set of candidate genes involved in floor plate differentiation were identified. Analysis of Foxa2 ChIP-seq dataset suggested that Foxa2 directly regulated more than 250 genes expressed by the floor plate and identified Rfx4 and Ascl1 as co-regulators of many floor plate genes. Experimental studies suggested a cooperative activator function for Foxa2 and Rfx4 and a suppressive role for Ascl1 in spatially constraining floor plate induction. Paper V addresses how time is measured during sequential specification of neurons from multipotent progenitor cells during the development of ventral hindbrain. An underlying timer circuitry which leads to the sequential generation of motor neurons and serotonergic neurons has been identified by integrating experimental and computational data modeling.