Efficient causal interaction learning with applications in microarray

Abstract

The prediction and identification of physical and genetic interactions from gene expression data is one of the most challenging tasks of modern functional genomics. Although various interaction analysis methods have been well studied in data mining and statistics fields, we face new challenges in applying these methods to the analysis of microarray data. In this paper, we investigate an enhanced constraint based approach for causal structure learning. We integrate with graphical gaussian modeling and use its independence graph as input of next phase's causal analysis. We also present graphical decomposition techniques to further improve the performance. The experimental results show that our enhanced method makes it feasible to explore causal interactions interactively for applications with a large number of variables (e.g., microarray data analysis). © Springer-Verlag Berlin Heidelberg 2005.