Protein expression data of The Cancer Genome Atlas

namely the Reverse Phase Protein Array data for statistical analysis

are adopted to mine hidden association information between cancer and genes

and to improve the efficiency of clinical diagnosis and to reduce the cost of inspection. Network structure of each group data and expression levels of different genes are gotten through the heat maps. And 5 genes which play an important role in protein expression levels for these 3 kinds of cancers are obtained by principal component analysis. Finally

the discriminant model based on the 5 genes for 3 kinds of cancer and the estimation of  the misjudgment rate are built by the back substitution method and cross-validation method. It is concluded that the network structure of the protein expression level for each kind of cancer is constructed respectively

5 genes which play an important role in protein expression level are sought out

and the result of linear discriminant model is reliable.