Netwerk gebaseerde omics data analyse

The omics revolution has introduced new challenges when studying interesting phenotypes. High throughput omics technologies such as next-generation sequencing and microarray technologies generate large sets of data for a single wet-lab experiment. Interpreting the resulting data from these experiments is not trivial due to the data size and the inherent noise of the underlying technologies. In addition to this, all these data lead to an ever expanding biological knowledge which has to be taken into account when analyzing new experimental results. Biological networks provide a useful and practical approach of representing this large amount of biological knowledge. Interaction networks for example provide a blueprint of biological pathways that can be activated in an organism under specific experimental conditions. These interaction networks provide an ideal representation to interpret high-throughput omics data and in addition to this, these networks can be used by computational methods to reconstruct the molecular mechanism that drives the specific phenotype under research. An illustration of using interaction networks to analyze and visualize the results of genetic screenings was performed in a first publication in the context of this thesis. To better understand the biological pathways that drive colony morphology in Saccharomyces cerevisiae, first an interaction network for this organism was constructed from publicly available interaction data and second the results of a genetic screening were mapped onto this network. Based on this analysis the biological pathways and molecular mechanism influencing colony morphology were identified and biologically corroborated. The main part of this thesis consists of the development and application of the PheNetic framework for subnetwork inference. Subnetwork inference is the computational reconstruction of the molecular mechanism responsible for an observed phenotype from interaction networks. Using high-throughput omics data, these methods “reason” about possible explanations or molecular mechanisms of how a specific phenotype works. In a first setup, i.e. proof-of-concept, the benefits of subnetwork inference methods and specifically PheNetic were illustrated. Using multiple differential expression data sets from knock—out experiments associated with reduced acid resistance in Escherichia coli, the molecular mechanisms underlying this phenotype could be identified and validated with literature data. From the results it became clear that subnetwork inference methods outperform naïve ranking of differentially expressed genes and as such the network methods could better identify the true molecular mechanisms that drive acid resistance in E. coli. A second setup was the interpretation of differential expression data using an improved version of the PheNetic framework. This application allows for the reconstruction of on the one hand the upstream regulatory network that induces the observed pattern of differential expression and on the other hand the activated downstream protein complexes and metabolic pathways in the observed phenotype. To provide a practical subnetwork inference tool that is readily applicable to experimental differential expression data a web server was developed available at http://bioinformatics.intec.ugent.be/phenetic/. In addition, the web server provides a visualization and analysis module for the interpretation of the inferrred subnetworks. A third setup was the identification of true “driver” mutations from experimental evolution experiments. Experimental evolution experiments induce a selection on an organism to adapt to an external stress, e.g. the presence of a toxic substance, limitation of nutrients, … . This type of experiments determines the genetic base of increasing fitness in a specific environment. By combining genetic data and differential expression data between the evolved and the original parent strain the molecular mechanisms associated with the increase in fitness can be identified. By assessing connectivity of the different mutations to the molecular mechanisms activated in the evolved strain, the true “driver” mutations, i.e. mutations that induce increased fitness, can be identified. This method was successfully applied on different evolution experiments in E. coli where previously known driver mutations could be identified from other mutations.status: publishe