Adilson Motter
Northwestern University
Genetic diseases, ecosystem collapses, cascading failures, network synchronization, materials design. These are some of the many outstanding interdisciplinary problems that could benefit from a predictive modeling approach to control the perturbation response of complex systems. The main obstacle to the development of such an approach has been that it is generally unclear how the large-scale collective behavior is affected by the local properties of the underlying interaction networks. In this talk I will discuss an alternative approach recently developed in my research group, which is based on inverting this perspective and seeking instead the conditions that should be imposed on the local network structure and/or dynamics to generate a desired (natural or human-selected) global collective behavior. In the context of cellular metabolism, this approach has been used to predict that a faulty or sub-optimally operating metabolic network can often be rescued by the targeted removal of enzyme-coding genes. Predictions go as far as to assert that certain gene deletions can restore the growth of otherwise nonviable gene-deficient cells, an effect now known as synthetic rescue. In food-web systems, it leads to the prediction that the removal or growth suppression of specific species can be used to mitigate the spread of extinction cascades. In oscillator networks, it explains once and for all why the common expectations that synchronization would be generally easier to achieve with more interactions, that synchronization properties would change monotonically as the number of available interactions is varied, and that certain network structures would facilitate while others inhibit synchronization, are all false. In the context of network design, this approach can be used to rationally design complex systems with new functional properties, such as mechanical networks that contract when tensioned or exhibit other forms of negative mechanical response to external forces. I thus hope to convey that, besides helping explain why "less can be more" in complex systems, these concepts promise to lead to a predictive modeling framework that can be used to control network response by only exploiting resources available in the system.
Friday, September 25th at 4:00 PM
Room L211, Technological Institute
Refreshments are served at 3:30 PM
Speakers Schedule
