The observation that genes rarely act independently has led to an interest in methods to analyze omics data in the context of gene interaction networks. These analyses are enabled by pathway databases that compile known biochemical interactions, which can then be combined with omics data to account for the complex correlation structure between genes and make inferences at the systems level. Network-based analyses of omic data can enable a mechanistic understanding of disease processes that cannot be gleaned from single-gene analysis. However, there is no consensus on the best way to integrate these sources of information, nor any clear criteria for evaluating the performance of network-based pathway analysis methods. I will describe our efforts toward establishing a rigorous framework to assess the reliability of pathway analysis methods. Our framework serves as a testing environment for the development of new methods and can guide users in choosing an analysis technique. I will also describe a novel graph-theoretic method for testing systems-level differences in the structure and function of gene regulatory networks, and demonstrate its performance relative to other methods.