Fuzzy C-means method for clustering microarray data

• Published in: Bioinformatics Vol. 19; no. 8; pp. 973 - 980
• Main Authors: Dembélé, Doulaye, Kastner, Philippe
• Format: Journal Article
• Language: English
• Published: Oxford Oxford University Press 22.05.2003; Oxford Publishing Limited (England); Oxford University Press (OUP)
• Subjects: Algorithms; Biochemistry, Molecular Biology; Biological and medical sciences; Cluster Analysis; Databases, Genetic; Fundamental and applied biological sciences. Psychology; Fuzzy Logic; Gene Expression Profiling - methods; Gene Expression Regulation - genetics; General aspects; Humans; Life Sciences; Mathematics in biology. Statistical analysis. Models. Metrology. Data processing in biology (general aspects); Neoplasms - genetics; Oligonucleotide Array Sequence Analysis - methods; Quality Control; Sequence Analysis, DNA - methods; Tumor Cells, Cultured; Yeasts; Yeasts - genetics; Yeast; Fuzzy algorithm; Classification; Empirical method; DNA chip; Gene expression; Web site; Bioinformatics
• ISSN: 1367-4803; 1460-2059; 1367-4811
• DOI: 10.1093/bioinformatics/btg119

Motivation: Clustering analysis of data from DNA microarray hybridization studies is essential for identifying biologically relevant groups of genes. Partitional clustering methods such as K-means or self-organizing maps assign each gene to a single cluster. However, these methods do not provide information about the influence of a given gene for the overall shape of clusters. Here we apply a fuzzy partitioning method, Fuzzy C-means (FCM), to attribute cluster membership values to genes. Results: A major problem in applying the FCM method for clustering microarray data is the choice of the fuzziness parameter m. We show that the commonly used value m = 2 is not appropriate for some data sets, and that optimal values for m vary widely from one data set to another. We propose an empirical method, based on the distribution of distances between genes in a given data set, to determine an adequate value for m. By setting threshold levels for the membership values, genes which are tigthly associated to a given cluster can be selected. Using a yeast cell cycle data set as an example, we show that this selection increases the overall biological significance of the genes within the cluster. Availability: Supplementary text and Matlab functions are available at http://www-igbmc.u-strasbg.fr/fcm/ Contact: doulaye@titus.u-strasbg.fr * To whom correspondence should be addressed.