A Clustering Method for Discovering Patterns Using Gene Regulatory Processes

Abstract

Clustering is a descriptive task that seeks to identify homogeneous groups of objects based on the values of their attributes (dimensions). The k-means and hierarchical as well as self-organizing maps have all been used for clustering expression profiles and a number of algorithms have been developed for expression data and applied to analyze it. These Clustering methods usually use metric distance for similarity measure. Correlation coefficient is also used but has a problem that it removes difference attributable to both the mean and the dispersion of the observations. Moreover, it may be unreasonable that every observation is assigned to one of clusters when the purpose is to find groups with similar pattern. Alter et al. [1] show that several significant eigengenes and the corresponding eigenarrays capture most of the expression information in field of genetics and some of the eigengenes represent independent regulatory programs or processes from its expression pattern across all arrays. Normalizing the data by filtering out the eigengenes (and the corresponding eigenarrays) that are inferred to represent noise or experimental artifacts enables meaningful comparison of the expression of different genes across different arrays in different expression. Such normalization may improve any further analysis of the expression data. Q-mode factor analysis has been used to find groups like clustering analysis and could be a good method to find patterns. However, this approach to clustering is plagued with a number of problems [3]. Genes with similar expression profiles may have something in common in their regulatory mechanisms. In this study, Q-mode factor analysis is used to model the gene regulatory processes which control genes and gene products and we modify the Q-mode factor analysis for discovering useful patterns in gene expression data. As a result of the factor modeling of gene expression data, our method can improve the result of clustering by removing noises and produce characteristic values of expression data.