We present preliminary results of a research project, in collaboration with Addenbrooke's Hospital and the University of Liege, that aims at studying cancer-related gene pathways by applying independent componenent analysis (ICA) techniques to gene expression data.
NA microarrays provide a huge amount of gene expression data and require therefore dimensionality reduction methods to extract meaningful biological information. ICA provides a means to tackle this task. Available ICA algorithms differ (i) by the contrast function they attempt to optimize in order to extract components that are "as independent as possible" and (ii) by the optimization scheme that they utilize to perform the optimization. In this work we investigate the efficacy of various contrast functions in the context of gene expression analysis. On the optimization side, we propose an innovative technique derived from differential geometric concepts.
Back to Control Seminars Page.