A Novel Local Features-Based Approach for Clustering Microarray Data

DNA Microarray technology makes it possible to moni- tor simultaneously the dynamic expression levels of tens of thousands of genes during some important biological pro- cesses. A first step to comprehend and interpret the result- ing mass of data is via clustering techniques. However, most existing methods are based on clustering genes by compar- ing their expression levels on all experiment conditions al- though genes in a functional cluster more often than not correlate only under a subset of conditions. Besides, most clustering algorithms depend on some critical user parame- ters in determining the number of resulting clusters. Unfor- tunately, correct parameter values are rarely known in real datasets. In this paper, we propose a novel clustering algo- rithm that (1) goes beyond global approaches to discovery gene clusters based on local features, and (2) automatically determines the number of resulting clusters. Furthermore, we introduce the norm-based method to improve it, as is proved reasonable. Extensive experiments are conducted on both synthetic and real data sets. Experiments prove that our method is efficiency and efficient.