A new clustering method for microarray data analysis

A novel clustering approach is introduced to overcome missing data and inconsistency of gene expression levels under different conditions in the stage of clustering. It is based on the so-called smooth score, which is defined for measuring the deviation of the expression level of a gene and the average expression level of all the genes involved under a condition. We present an efficient greedy algorithm for finding clusters with a smooth score below a threshold after studying its computational complexity. The algorithm was tested intensively on random matrices and yeast data. It was shown to perform it well in finding co-regulation patterns in a test with the yeast data.