| With the development of methylation microarray technology, large amount of DNAmethylation data of different cancer types are accumulated. Aberrant methylation ofgene promoters plays important roles in carcinogenesis. However, the functionalsignificance of aberrant methylation of gene promoters in carcinogenesis currentlyremains unclear. Also, a large number of genes are differentially methylated in cancergenome, how to extract the “driver†genes is an important task. Thus, we systematicallyanalyzed the functional roles of the genome-wide hypermethylation andhypomethylation of gene promoters in carcinogenesis. Then, we developed an approachfor integrating promoter methylation and gene expression data to identify “driver†genesof cancer, and finally did some adjustment for analyzing the paired cancer-normalsamples.The main contributions are as follows:1. Functional patterns of promoter hypomethylation and hypermethylation incancer genomes. Based on genome-wide methylation data for five different cancer types,we proved that genes with promoter hypermethylation were highly consistent infunction across the five cancer types, and so were genes with promoter hypomethylation.Specific functions related to developmental process, regulation of biology process,signal transduction within the cell and behavior of an organism in response to externalor internal stimuli were significantly enriched with hypermethylated genes whereasdepleted of hypomethylated genes. In contrast, certain functions related to cell killingand response to stimulus, including immune and inflammatory response, wereassociated with an enrichment of hypomethylated genes and depletion ofhypermethylated genes. On one hand, these results show that genes with promoterhypermethylation and genes with promoter hypomethylation are highly consistent infunction across different cancer types, respectively. On the other hand, hypermethylatedand hypomethylated genes in cancer tend to disturb different functions.2. Uncover “driver†genes in cancer. Based on the assumption that methylationaberration of a “driver†gene may lead to expression alternation of itself and a set of genes associated with cancer pathways, we developed a computational framework forintegrating promoter methylation and gene expression data to identify “driver†genes ofcancer. Applying this approach to three breast cancer data, we identified many novelbreast cancer “driver†genes. With known cancer gene database and protein-proteininteraction network, we evaluated the reliability of these novel “driver†genes, and ourapproach proved effective in identifying cancer “driver†genes. At last, we found thatsome of the identified driver genes were subtype-specific for basal-like, luminal-A andHER2+subtypes of breast cancer.3. Uncover “driver†genes in paired cancer-normal datasets. Taking the advantagethat data of paired samples could provide the relative methylation change informationfrom normal to tumor for each individual patient, we did some adjustment to the methodbefore and developed an approach to identify “driver†genes with integrated promotermethylation and gene expression data generated from paired cancer and normal samplesfor each of a cohort of breast cancer patient. We applied this approach to analyzeintegrated samples of five breast cancer datasets and discovered some novel cancerdriver genes. With known cancer gene database, we evaluated the reliability of thesenovel “driver†genes and our approach also proved effective in identifying cancer“driver†genes.In summary, we systematically analyzed the functional roles of the genome-widehypermethylation and hypomethylation of gene promoters in carcinogenesis, and provethat the hypomethylation genes are as important as hypermethylated genes. At the sametime, we developed an approach for integrating promoter methylation and geneexpression data to identify “driver†genes, which a fundamental step towardsmechanistic characterization of cancer and may provide potential targets of epigenetictherapy. |
PDF Full Text Request