Bioinformatics Seminar - Tuesday, Oct. 2 at 4:30PM
October 2 @ 4:30 PM - 5:30 PM - Lily G126
Network motif identification and structure detection in biological databases using graphical models
Munni Begum, Department of Mathematical Sciences, Ball State University, Muncie, IN
Tuesday, October 2, 2012
Identification and structural analysis of smaller, more localized networks of interactions, or motifs, contribute significantly to our overall understanding of the underlying biological processes. This leads to discoveries in the fields of personalized medicine, microbiological industrial processing, and others. In this work, we explore identification of network motifs, small sub-networks that appear in a network more often than statistically expected. We employ rich social network methodologies such as p* models in order to identify statistically significant local features of a given biological network. In particular, our work extends available methodologies based on exponential random graph models (see for example, Saul and Filkov, 2007, doi: 1093/bioinformatic /btm370) in multiple ways. We explore additional local network features; and we study applications to a broader range of databases. Implementation of computational methods, such as, Metropolis-Hastings algorithm and!
other Bayesian methods, allows us to obtain efficient estimates of model parameters. These methods are applied to commonly known biological databases including Escherichia coli from RegulonDB and metabolic networks from the WIT database. Results are compared in terms of how well these methods fit to observed network data.
Zachary M. Saul* and Vladimir Filkov. 2007. Exploring biological network structure using exponential random graph models. Bioinformatics. Vol. 23 no. 19 2007, pages 2604-2611.