Seminar 3rd March 2010 2 p.m. B6/1083
Cluster models for joint analysis of -omic data
Dr. Simon Rogers
University of Glasgow
- Web page
- http://www.isis.ecs.soton.ac.uk/seminars/?action=viewpresentation&presentation_id=486
- Categories
- Systems biology
- Submitter
- Richard Edwards
Clustering is popular with biologists. Many different clustering approaches have been applied to large biological datasets to, for example, find groups of genes that behave similarly (suggesting some shared functionality). Often we might have more than one dataset available to us. For example, I will talk about a particular dataset that provides (time-series) measurements of mRNA and protein levels for the same set of genes. Analysing the two datasets together has the potential to provide more explanatory power than analysing them individually. The most obvious approach is to just stick the two datasets together and cluster the result. In this talk, I will show how for this particular dataset this isn't very sensible and will present two cluster models that attempt to analyse the two datasets together in a more flexible way. Both methods are based on statistical mixture models. Although I'm using biological data, it is possible that there are many other interesting applications for techniques such as these.