Complex networks of time-series: what does it reveal more than local interactions?

In many complex systems, the only observable variables are the time-series and event-series of their components. There is a huge literature about extracting the interaction network of these systems in molecular biology, neuroscience and economy. Although this approach invigorates these disciplines to deal with large data, it usually focuses on microscopic results. In this presentation, I will suggest some holistic approaches towards the analysis of these networks, based on two examples: medical words network evolution and stock market network near the crisis. Finally, I will try to connect the measured global indicators to dynamics of the system, using the idea of symmetry breaking in the spin glass models.