Large-scale Bayesian networks

Large-scale gene expression data from DNA microarrays or RNA sequencing often include hidden patterns of correlation between genes. These patterns reflect signaling and regulatory events and other complex molecular processes underlying cellular metabolism and physiology. Several gene network modeling techniques have been applied in order to infer meaningful biological networks from large-scale genomic data.  Of these, Bayesian Networks (BN) methodologies are especially appealing, as they provide a framework for incorporating prior knowledge. Prior knowledge is available through the many large and steady-growing repositories of molecular interactions and relations, such as protein-protein interactions, transcription factor binding data or protein sequence homologies. However, since the introduction of BN for genetic data by Friedman in 2000, only a few methods making use of prior knowledge within the BN framework have been proposed.

The candidate should take part in formulating and implementing how to use prior information within Bayesian Networks. Our group has already developed a method for incorporating prior information to find clusters of correlated genes from large-scale expression data. In this project we seek to infer directly connected genes within these clusters. The candidate will work together with people with competence in mathematical statistics, programming and molecular biology.

The candidate should be skilled in programming and mathematical statistics. Special competence in Bayesian statistics is an advantage, but can be attained during the work.

Presentation: Building large-scale Bayesian networks using informative priors

Publisert 19. okt. 2011 14:30 - Sist endret 22. okt. 2013 12:56

Omfang (studiepoeng)