Lecturer: Gayle Leen, PhD

Course assistant: José Caldas

Topic at autumn 2009: "Advanced probabilistic models in bioinformatics".

First meeting on Thursday, 24th September 2009, at 12:15 in A328.

Course Overview: 

High throughput technologies have completely transformed the concept of data in bioinformatics; collections of data
measurements from different biological systems of interest tend to be vast, heterogeneous, and contain complex interactions between the system variables. There has been increasing interest in developing probabilistic approaches to analysing biological data since they are a principled way of dealing with noise / uncertainty, can allow the incorporation of prior knowledge / structure about the problem domain, and may lead to better interpretability of biological results.

In the course we will look at different aspects of functional genomics problems: how do genes function together within the entire human genome? what mechanisms are used? which genes are functionally related? how does gene function change between different conditions (e.g. cancer and non-cancer)? and then look at current, advanced probabilistic modeling techniques for addressing these problems.
Clustering gene expression data can find different classes of genes behaving in the same way, assuming similar expression patterns suggest co-regulation. Model based approaches can make the assumptions about the problem structure more explicit and uncover functional relationships and dependencies underlying the global expression process. We will investigate how the idea of gene function / functional modules can be formulated and inferred from the data in this framework, as complex probabilistic models. Possible topics include Bayesian automatic relevance determination algorithms, model based clustering, hierarchical, nonparametric Bayesian methods, probabilistic integration of auxiliary sources of data.