Clustering dependencies with support vectors

Experimental technologies in molecular biology (particularly oligonucleotide and cDNA arrays) now make it possible to simultaneously measure mRna-levels for thousand of genes [1]. One drawback is the difficulty in organizing this huge amount of data in functional structures (for instance, in cluster configurations); this can be useful to gain insight into regulatory processes or even for a statistical dimensionality reduction. In this chapter, we consider a simplified model based on mRna-data only, which is an effective gene-to-gene interaction structure. This can provide at least a starting point for hypotheses generation for further data mining. The chapter is organized as follows. In Sect. 11.2 we givea brief overview of the kernel methods and SVC algorithm. In Sect. 11.3 we address the MGRN problem and in Sect. 11.4 we apply our formulation to clustering the training set. In Sect. 11.5 we discuss the numerical results and finally, in Sect. 11.6, we conclude and discuss some directions for future work. © 2008 Springer Science+Business Media, LLC.