Dynamic Clustering of High-Dimensional Biological Data

• Activity Number: 271
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Rebecka Jornsten*+ and Jose Sanchez Lopez
• Companies: Chalmers/Gothenburg University and Chalmers/Gothenburg University
• Keywords: clustering ; gene expression ; variable selection

Abstract

With the data deluge in the biological sciences, clustering has become an ever-increasingly popular approach to summarize data. To analyze microarray gene expression clustering, we commonly group genes that exhibit similar expression profiles across experimental conditions. In the work we present here, we view clustering as a dynamic problem. We thus define an evolving cluster structure that alters in terms of gene membership as experimental conditions are included or excluded. The statistical problem we focus on is one where an ordering of experimental conditions are provided according to the users preferences, such as the greatest stability of partitions in response to experimental conditions being added to the analysis, or an ordering with the most dynamic clustering results. We interpret stable and dynamic gene sets that this clustering methodology provides on real application data.