Statistical methods in graphs with applications to Life Sciences
Abstract
The importance of statistics in natural sciences is unquestionable. Statistics is essential to analyze data appropriately and to reach reliable conclusions. However, little is known about formal statistical methods on graphs and their theoretical properties even with an increasing number of reports on the analysis of real world networks (e.g. functional brain networks, protein-protein interaction networks, and social interaction networks). Networks are usually analyzed using computational algorithms based on graph theory, such as calculation of centrality measures (relative importance of vertices and edges) or identification of structural patterns (motifs). The main drawback of this approach is the fact that real world networks present intrinsic fluctuations (random noise) that are not taken into account by these "classical" algorithms. Therefore, methods with statistical perspective may aid and complement these analyses. The main goals of this proposal is the development of both theory and computational statistics methods and apply them to real world data, such as networks from molecular biology, neuroimaging, and cardiac data. The development of this project will be essential to obtain novel insights, solidify the cooperation among PIs, and improve the research quality of all involved groups. In the long term, we will consolidate the field of Network Statistics, form groups of highly qualified researchers, and ultimately build a Center for Network Statistics in Brazil. This center will develop novel theoretical frameworks, methodological tools and also apply the latter to solve health sciences problems. (AU)