Understanding the cellular behavior from a systems perspective requires the identification of functional and physical interactions among diverse molecular entities in a cell (i.e., DNA/RNA, proteins, metabolites,...). A network-based representation captures many of the essential characteristics of these various biological systems and can be exploited to study a molecular entity like a protein in a wider context than just in isolation and can provide valuable insights into the systems mode of action and functionalities. Unraveling these molecular networks is considered one of the foremost challenges in current bioinformatics research.Powerful and scalable technologies enabled the generation of genome-wide datasets that describe cellular systems by capturing the interactions of their building blocks or by characterizing the state of a system under different environmental stimuli. The distinct nature of these datasets often brings about complementary views on cellular behavior and integration of them holds the key to the successful reconstruction of the underlying networks. In a first part of the thesis we therefore aim to infer networks from diverse of omics datasets. We present ProBic, a method to identify modules of genes that show co-expression to a set of genes of interest (i.e., query or seed genes), and the conditions in which they are co-expressed. These modules are termed biclusters, and represent potentially co-regulated genes, highlighting part of the transcriptional network. We applied ProBic on a benchmark set in E. coli and showed that high quality biclusters with biological relevance could be obtained.Next, we integrated information from several functional datasets to predict protein-protein interactions from public experimental interaction data, using a naive Bayesian classifier. Clustering the obtained protein network illustrated the presence of functionally coherent modules, and showed the opportunity of assigning novel gene functions based on cluster functionality. Viewing a single entity or an experimental dataset in the light of an interaction network can reveal previous unknown insights in biological processes or functional behavior. Methodologies that identify and explore paths in networks between given input and output nodes have gained much interest. Such a path in a network can be seen as a mechanistic representation of the way information propagates through the network. Identifying biologically meaningful paths in the network between nodes of interest, nodes which can be defined from functional datasets that are independent from the network itself, can unveil previously uncovered signal flow mechanisms that are responsible for the observed functional behavior or define a measure for relatedness of two nodes in the network.In a second part of the thesis we present a novel network-based method to interpret eQTL data. One of the challenges in eQTL analysis, besides the identification of genetic loci that are associated to gene expression, is the exact identification of the true causal gene in an associated locus, causing the variation in gene expression. The method therefore aims to prioritize candidate genes in a locus based on their network-relatedness to a set of associated target genes. We show that the approach outperforms a state-of-the-art gene prioritization method in case a common causal factor is present for a set of targets, based on several performance criteria.Finally, we applied the methodology on a biological dataset in yeast that combined both genomic and expression variation of a pool of yeast segregants. We predicted several genes to be involved in ethanol production capacity, which is an important phenotype in the fermentation industry.