Identification of oncogenic genes for colon adenocarcinoma from genomics data

Identification of oncogenic genes from comprehensive genomics data with large sample size is of challenge. Here, we apply a well-established computational model, Bayesian factor and regression model (BFRM), to predict unknown colon cancer genes from colon adenocarcinoma genomic data. The BFRM takes advantages of its latent factors to characterize the underlying association between genes and the large number of colon cancer patients. Based on the known cancer genes in Online Mendelian Inheritance in Man (OMIM), we addressed three important latent factors focusing on characterization of heterogeneity of expression patterns related to specific oncogenic genes from the microarray data of 174 colon cancer patients. We found that the three latent factors can be employed to predict unknown colon cancer genes using the known oncogenic genes. These predicted unknown cancer genes were extensively validated by using the new somatic genes identified in the same patients from DNA sequencing data.