Pathway modeling: From gene expression to pathway dynamics

Biological pathways represent a critical level of biological organization and understanding of biochemical pathways is identified as key to future advances in biological sciences (Schaefer, 2004). The overall goal of this thesis is to develop a pathway based approach that integrates different aspects of biological pathways, specifically the structure and the dynamics of a pathway in order to characterize cells' behavior. Our objectives are to assess structural and functional cross-species comparison of pathways (Chapter 2), to formulate a reliable pathway activity metric based on gene expression data (Chapter 3), to demonstrate that our pathway activity formulation can predict the underlying dynamics (Chapter 4) and finally to demonstrate that the pathway activity formulation can identify cell's response to a stimulant (Chapter 5). Chapter 3-5 present how a significant pathway can be identified. Then, cross-species comparison of pathways (Chapter 2) can be used. Note that we could have Chapter 2 and Chapter 5 swapped for a more fluent flow. Neverthless, we present the chapters in this order for a better read.;In Chapter 2, we propose an improvement of the reaction alignment method, emerged as the most successful pathway comparison method, by accounting for sequence similarity in addition to reaction alignment method. Using nine species, including human and some model organisms and test species, we evaluate the standard and improved comparison methods by analyzing glycolysis and citrate cycle pathways conservation. In addition, we demonstrate how organism comparison can be conducted by accounting for the cumulative information retrieved from nine pathways in central metabolism as well as a more complete study involving 36 pathways common in all nine species. In Chapter 3, we explore an extension of the pathway activity methodology which entails singular value decomposition (SVD) of the expression data of the genes constituting a given pathway. We show that pathway analysis enhances our ability to detect relevant changes in pathway activity using synthetic data. In addition, we illustrate that pathway activity formulation should be coupled with a significance analysis to distinguish significant information from random deviations. In Chapter 4, we perform an unsupervised pathway level analysis, based on the formulation presented in Chapter 3, on a rich time series of transcriptional profiling in rat liver. The over-represented five specific patterns of pathway activity levels, which cannot be explained by random events, exhibit circadian rhythms. The identification of the circadian signatures at the pathway level identify pathways related to energy metabolism, amino acid metabolism, lipid metabolism and DNA replication and protein synthesis, which are biologically relevant in rat liver. In Chapter 5, we demonstrate that our pathway activity formulation enables us to detect relevant changes in pathways due to in utero di-butyl-phthalate (DBP) exposure. Our findings suggest that the pathways that produce precursors to cholesterol synthesis exhibit more significant change compared to the rest of the affected pathways. In addition, pathway activity levels of certain biological functions accompany testosterone decrease, which is the critical event for male reproductive developmental effects of DBP, such as steroid hormone metabolism and biosynthesis of steroids.