A tool for cluster number estimation in SOM-based gene expression pattern análisis

FERNÁNDEZ, ELMER ANDRÉS; BALZARINI, MÓNICA

Angra Dos Reis, Brasil

Simposio; Brasilian Symposium on Bioinformatics; 2007

Abstract. Cluster methods are crucial to study genomic patterns of coexpressedgenes. Neural network algorithms such as self organizing map (SOM) havebeen extensively used to cluster gene expression data. Result visualizationtechniques are important tools for cluster recognition in SOM. In this work asimple tool that implements an algorithm to identify and visualize clusters isproposed. It is based on two concepts (Relative Position and Q statistics) thatcan be applied to a SOM network. The Relative Position is a new SOM nodeadaptiveattribute defined from the node moving within a two dimensionalspace imitating the movement of the SOM codebook vectors in the input space.By means of the Q statistics the algorithm evaluates the SOM structureproviding an estimate of the number of clusters underlying the data. The toolallows the visualization of the cluster node patterns facilitating clusterinterpretation.