Gene cluster algorithm based on most similarity tree

As the development of DNA array technology, large-scale DNA array expression data sets are produced. It is very important to construct the functional genome and denote the functions of unknown genes. This manuscript describes a gene cluster method based on the most similarity tree (CMST), which is a partition of equivalence groups of equivalence relation with similarity measure. The Gap statistic of similarity measure is introduced to determine the most optimal similarity measure and an optimally self-adaptive gene cluster algorithm based on CMST (OS-CMST) is proposed. The cluster method of CMST can get the global optimal clusters and the experiment results show that CMST outperform traditional cluster methods of K-means and SOM