UNIVERSITY OF CALIFORNIA Los Angeles Study of Functionally Related Gene Groups Using Microarray Expression Data: Theory and Application

Various clustering methods have been applied to microarray gene expression data to identify genes with similar expression profiles. Recently, as the biological annotation data accumulated, many genes have been organized into functional categories such as Gene Ontology. Because functionally related genes may be regulated by common cellular signals, and hence be co-expressed, how to utilize these rapidly increasing functional annotation resources to improve the performance of clustering methods is of great interest. In addition, some “scattered” genes may have distinct expression profiles and may not co-express with other genes. Identification of these “scattered” genes could further enhance the performance of clustering methods. We developed a new clustering algorithm, Dynamic Weighted Clustering with Noise set (DWCN), which makes use of the prior gene annotation information and allows for a set of scattered genes (“noise”) be left out of the main clusters. This could significantly improve the quality of the final clusters especially when a considerable percent of such noisy genes is present in the data. An application in a yeast cell-cycle gene expression data demonstrated that our method could produces clusters with more consistent functional annotation as well as more coherent expression pattern than traditional clustering techniques.