@ARTICLE{, author = {}, title = {Gene Expression Data Clustering with Random Forest Dissimilarity}, volume = {22}, number = {136}, abstract ={Background: The clustering of gene expression data plays an important role in the diagnosis and treatment of cancer. These kinds of data are typically involve in a large number of variables (genes), in comparison with number of samples (patients). Many clustering methods have been built based on the dissimilarity among observations that are calculated by a distance function. As increasing the dimensions reduces the performance of distance functions, most of the methods provide low accuracy. In this paper a new dissimilarity measure is introduced based on a classification method, called Random forests (RF). The performance of this new measure has been evaluated in the gene expression data. Methods: In this article, the clustering problem of Chowdary data set, using the RF dissimilarity measure, is under consideration. At the first step, the clustering problem is converted to classification problem, thereafter the new dissimilarity is calculated using the classification method of random forests. Finally, the data are clustered with a partition around mediod algorithm and the results are then evaluated by adjusted rand index. All the analysis is implemented with R software. Results: The value of adjusted rand index (0.8149) represents an acceptable agreement between clusters and true groups. The most effective gene in constructing the clusters was gene no.31 which was detected by using the unique ability of RF that is identifying the importance of variables. Conclusion: The random forest dissimilarity is an efficient criterion for measuring dissimilarity in gene expression data clustering. Detection of effective genes in clustering that is done with RF, helps the researcher in the diagnosing and treatment of the cancers }, URL = {http://rjms.iums.ac.ir/article-1-4097-en.html}, eprint = {http://rjms.iums.ac.ir/article-1-4097-en.pdf}, journal = {Razi Journal of Medical Sciences}, doi = {}, year = {2015} }