Since the 20th century,cancer has become one of the major hazards to human health,the tumor has become a global medical problem; as we all know,tumor is a kind of disease which is closely related to the genetic mutation,and with the completion of Human Genome Project(HGP),leads to tumor genome research raised.The purpose of tumor genome research is to find out the primary mutations that result in the development and progression of tumor,find out driver genes in tumor.In this study,we are on the basis of the studies of “duons” codon and tumor genome to explore the relationship between “duons” codons and tumor driver genes.Here,we detect the somatic mutations and INDELs of 5325 tumor samples representing 22 tumor types,and used MutSig software to identify genes that were mutated more often than expected by chance given background mutation processes by analyzes lists of mutations discovered in sequencing data and find out driver genes in these tumor types.Then,we combined the information of driver genes codon mutations with the frequency distribution of “duons” codon in coding regions to explore the relationship between “duons” codons and tumor driver genes.After a preliminary analysis,we found that the mutations of most of tumor were not all in codons which have high “duons” frequency,messy distribution.But in CpG region,we can find many mutations,so we realized that the relationship of encoding protein codons preferences and mutation.