Learning time-varying interaction networks

Licentiate thesis, 2013

Most biological systems consist of several subcomponents which interact with each other. These interactions govern the overall behaviour of the system; and in turn vary over time and in response to internal and external stress during the course of an experiment. Identifying such time-varying networks promises new insight into transient interactions and their role in the biological process. Traditional methods have focussed on identifying a single interaction network based on time series data, ignoring the dynamic rewiring of the underlying network. This thesis studies the problem of inferring time-varying interactions in gene interaction networks based on gene microarray expression data. With the advent of next generation sequencing techologies, the amount of publicly available microarray expression data as well as other omics data has grown tremendously. Further, the microarray data is often generated from different experimental conditions or under network perturbations. One of the current challenges in systems biology is integration of data generated from different experimental conditions and under different stresses towards understanding of the dynamic interactome. NETGEM, the first study included in this thesis describes a method for inference of time-varying gene interaction network based on microarray expression data under network perturbation. The method presents a probabilistic generative model under the assumption that the changes in the interaction network are caused by the changing functional roles of the interaction genes during the course of a biological process. This is used to infer time-varying interactions for a perturbation study in {\em Saccharomyces cerevisiae\/}~(Baker's Yeast) under nutrient stress. The inferred network agrees with experimental evidence as well as identifying key transient interactions during the course of the experiment. In the subsequent study, we present a survey chapter describing current approaches for inference of time-varying biological networks based on node observations. We give an overview of different methods in terms of the underlying model assumptions and applicability under different conditions. We also describe how recent advances in theory of compressed sensing have led to development of new network inference methods with mild assumptions on network dynamics.