Networks have been used to model a variety of real-world phenomena in many domains. Due to limitations of techniques for data collection, traditional network research has typically focused on studying static and homogenous networks. However, many interactions (e.g., social communications or relationships between biomolecules) are evolving and vary in type. With the recent advancement of data collection techniques, increasing amounts of dynamic and heterogeneous network data are becoming available. Extracting knowledge from such data is a non-trivial task due to the lack of methods for their analyses and consequently many challenging questions have emerged both on the computational as well as the application side.
Therefore, this Ph.D. dissertation focuses on developing computational strategies for analyzing dynamic and heterogeneous networks and studying their interdisciplinary implications. Here, we explore the domains of social and biological networks, although the strategies are applicable to other domains as well. In particular, we are interested in three key questions: 1) Will studying data via heterogeneous network analysis result in different findings compared to studying the same data via homogeneous network analysis? 2) How to systematically analyze network data that is both heterogeneous and dynamic? 3) How to efficiently compare two heterogeneous yet related networks via network alignment?
To this end, we: 1) integrate heterogenous network data and demonstrate that our approach reveals additional information that is missed by simpler approaches such as homogenous network analysis, by exploring a smartphone study encompassing multiple link types and node traits; 2) introduce a novel computational framework for systematic analysis of dynamic and heterogeneous networks, which we use to link individuals’ evolving social network positions with their traits, revealing in the process additional links that are missed by simpler approaches such as static network analysis or that have not been studied to date; and 3) introduce the first ever comparison of two complementary types of network alignment methods (local and global) and propose a new algorithm, IGLOO (Integrating Global and LOcal biOlogical network alignment), to reconcile the two, demonstrating in the process the superiority of IGLOO over each network alignment type individually.