A Markov chain model-based method for cancer classification

In this paper, we propose a Markov chain model (MCM)-based method for cancer classification. By viewing a gene chain extracted from a gene pathway map as a gene Markov chain (GMC), the method can construct a MCM for different cancer classes. The resulted MCM captures the co-activity pattern of genes in terms of an initial state distribution and a state transition probability matrix. When used for cancer classification, the method first calculates the respective probabilities of a test sample belonging to different cancer classes according to the corresponding MCM, and then predicts the class of the sample as the one with the highest probability. We evaluate the proposed method on the publicly available leukemia dataset and compare it with several conventional methods.