Identifying Characteristic Genes Based on Robust Principal Component Analysis

In this paper, based on robust PCA, a novel method of characteristic genes identification is proposed. In our method, the differentially expressed genes and non-differentially expressed genes are treated as perturbation signals S 0 and low-rank matrix A 0, respectively, which can be recovered from the gene expression data using robust PCA. The scheme to identify the characteristic genes is as following. Firstly, the matrix S 0 of perturbation signals is discovered from gene expression data matrix D by using robust PCA. Secondly, the characteristic genes are selected according to matrix S 0. Finally, the characteristic genes are checked by the tool of Gene Ontology. The experimental results show that our method is efficient and effective.