Abstract: The relativity of protein-protein interaction ( PPI) can be reflected by the gene coexpression correlation． The study of global PPIs in cancer cells or tissues will facilitate the understanding of the relationship between geno- types and phenotypes，as well as the mechanisms of cancer development in normal and abnormal cells or tissues． The mRNA expression profile data and clinical information was collected from the Cancer Genome Atlas ( TCGA) ，including 5726 samples across 11 human cancers type． Combining with PPI gene pairs of Human from STRING and HPRD databases，Pearson correlation coefficients ( PCC) was used to calculate the gene coexpression coefficients of tissue-specific PPIs in normal samples and cancer samples at different stages，and then the cancer-disordered and cancer-specific PPI gene pairs were obtained． With the different gene expression and survival analysis，PPI gene pairs which may affect the development of cancer were found out． Result shows that: the gene coexpression coeffi- cient of PPI gene pairs in cancer samples is significantly lower than normal samples． There are 2812 pairs of can- cer-disordered PPI genes，among which LRSAM1，ATXN1，SMAＲCC2 and SMARCA2 appear in multiple cancers interacted and clustered into a network module，and they may play an important role in caner． There are 31 pairs of cancer-specific PPI genes，among which BMP1 and COL1A1 may interact to promote tumorous migration and infil- tration． In addition，113 pairs of PPI genes with differential expression were selected from cancer-disordered and cancer-specific genes． Among them，16 pairs show significantly survival difference in 7 cancers．

Key words: cancer, TCGA, protein-protein interaction, gene coexpression