Stability Evaluation of Clustering Algorithms for Time Series Gene Expression Data

Different clustering techniques such as Self-Organizing Map (SOM), and hierarchical clustering, among others, have been applied to gene expression data. The focuses of theses studies are often on the biological results, and there is no indication on what methods are more suitable for clustering gene expression. In this paper, an evaluation methodology that assesses the stability of clustering methods in relation to external validation criteria is presented. In order to do so, an investigation of the extent to which similar results are obtained from analyses of subsets of the data is made. In this paper, preliminary results of applying this methodology to evaluate the stability of three clustering techniques (SOM, hierarchical clustering, and dynamical clustering) in relation to five gene expression time series data sets from Yeast cells are shown.