Wavelet Packet Decomposition-Based Fuzzy Clustering Algorithm for Gene Expression Data

Abstract

Analysis of gene expression data is the base of gene regulatory relations and gene network model construction, which has become the hotspot of bioinformatics. Many clustering methods have been applied to analyze gene expression data. Clustering results may be affected by inherent noise in gene expression data. A wavelet packet decomposition based denoise scheme of gene expression data analysis is put forward in this paper. The principles and methods of wavelet packet decomposition, best wavelet packet bases selection and threshold selection are discussed in detail. Fuzzy c-means clustering method is introduced. Clustering results are listed both for original gene expression data and wavelet packet decomposition processed gene expression data. Comparison of clustering results shows that the method can reduce background noise effectively. Clustering accuracy after wavelet packet decomposition is higher than that of direct clustering and can detect more precise categorized information, which also corresponds with our speculation