From biological networks to social networks, we are surrounded by manycomplex networks with various nature and structure. Given the important rolethese networks, and many others, play in our daily life, in science and ineconomy, their understanding, mathematical description, prediction, and controlhave become a major intellectual and scientific challenge of the 21st century.In response, the field of Network Science has emerged, drawing ontheories and methods including graph theory, statistical mechanics, datamining, inferential modeling, and social structure. In this dissertation, wefocus on the application of Network Science in two distinct areas: network structureand health.
In the first chapter, we explored the properties of physical networks,where the nodes and links are physical objects unable to cross each other.These non-crossing conditions constrain their layout geometry and affect hownetworks form, evolve and function. We developed a modeling framework thataccounts for the physical reality of nodes and links, allowing us to explorehow the non-crossing conditions affect the network geometry. For small linkthicknesses, we observed a weakly interacting phase where the layout avoids thelink crossings via local link rearrangements, without altering the overalllayout geometry. Once the link thickness exceeds a critical threshold, astrongly interacting phase emerges, where multiple geometric quantities scalewith link thickness. We observed a deep universality, finding that the observedscaling properties are independent of the underlying network topology.
In the second chapter, we investigated the role that diet plays in thedevelopment of Coronary Heart Disease (CHD). We applied an Environment-WideAssociation Study (EWAS) approach to Nurses’ Health Study data to explorecomprehensively and agnostically the effect of 257 nutrients and 117 foods onCHD risk. Our implementation of EWAS successfully reproduced prior knowledge indiet-CHD associations and helped us detect new associations that werepreviously only poorly studied in the literature. We showed that EWAS allows usto unveil the bipartite food-nutrients network, highlighting which nutrient inwhich food drives CHD risk. We showed that there is a distinct clustering insuch network where protective nutrients and foods are highly interconnected inone cluster, so do harmful nutrients and foods in another. Using this network,we showed that solely looking at food items, one would underestimate the effectof those nutrients whose consumption is strongly determined by the behavioralaspect and not mainly by their average amount in food.
Dissertation Committee Members:
Professor, Network Science Institute, Northeastern University
Ozlem Ergun(Committee Member)
Professor, Department of Mechanical and Industrial Engineering, Northeastern University
Mehdi Behroozi(Committee Member)
Assistant Professor, Department of Mechanical and Industrial Engineering, NortheasternUniversity