Multiscale Network Reconstruction from Gene Expression Measurements: Correlations, Perturbations, and A Priori Biological Knowledge

In the last decade, physics has been expanding to new research areas. In particular, life-related sciences (ecology, sociology, economics, and last but not least biology) have been showing striking analogies with complex systems arising from various physical areas. Such an approach has happened from both fronts: on the life sciences side, huge amounts of data have become available for detailed analysis, thanks also to the Internet, through which these data are nowadays easily collectable and queryable (e.g., stock market financial series, tables of social relationships from movie copart- nerships to e-mail fluxes, high-throughput biological data). On the other side, many physical and mathematical tools that had proven useful in explaining complex phenomena like polymer growth or spin glasses began to spread to other research areas like biological and social sciences in a broad sense. The common trait of these research fields can be found in the framework of network theory, such that focusing on the relationships among elements allows us to draw general conclusions even though the details of the system are not completely known or easily tractable from a mathematical point of view. Relaxing attention to the details of the specific interaction or element, network theory aims to provide tools for the characterization of a set of relationships, represented as edges or links, occurring among similar elements, referred to as vertices or nodes.