Tumor Clustering Based on Penalized Matrix Decomposition

A reliable and precise identification of the type of tumors is essential for effective treatment of cancer. In this paper, we proposed a novel method to cluster tumors using gene expression data. In this method, we use penalized matrix decomposition (PMD) to extract metasamples from gene expression data. Specially, a metasample can capture structures inherent in the samples in one class. In addition, we present how to use the factors of PMD to cluster the samples. Compared with traditional methods, such as HC, SOM and NMF etc., our method can find the samples with complex classes. At the same time, the number of clusters can be determined automatically.