Title: Graphical and Network Models in Bioinformatics
Speaker: Sujay Datta (Associate Professor, University of Akron, Statistics Department)
Abstract: In recent years, graphical and network models have become increasingly useful in certain areas of the biomedical sciences including genomics, proteomics, genetic epidemiology and systems biology. This has been facilitated by a number of new developments in structure learning algorithms and their evolution from constraint-based to score-based and hybrid algorithms. Examples include Gaussian graphical models, multivariate Bernoulli graphical models, hypergraph models, Bayesian networks and dynamic Bayesian networks. Also, motivated by the increasing popularity of such models, new methodologies have emerged in the statistical literature enabling us to do statistical inference with various aspects of them. Methods have recently been developed for parametric and bootstrap-based assessment of variability in the learned network structures, introduction of the notion of ‘confidence’ to structure learning, incorporation of graphical constraints into penalized regression and incorporation of measurement errors into structure learning. In this presentation, we delve into the theory and methodologies of graphical and network modeling in the contexts myogenic progenitor differentiation and protein-protein interaction.