Abstract
When a comparative analysis of gene expression data is conducted, the clustering method has been extensively used for finding out meaningful associations between different gene clusters. With the aim of obtaining a high quality of associations between different gene clusters without any inconsistency, we developed a novel method that is composed of the following three major steps; 1) to remove abnormal measurements, 2) to apply the K-Means to all the samples and to obtain the output of graphic patterns, and 3) finally to identify reasonable associations among different gene clusters. In this presentation, we present this new method and discuss its applicability to actual data.