DNA Methylation Module Network-based Prognosis And Molecular Typing Of Cancer

Achieving cancer prognosis and molecular typing is critical for cancer treatment.Prognosis usually means the estimate of success with treatment and chances of recovery,which is closely related to the molecular typing of cancer.As a complex polygenic disease,the same cancer performs differently in different individuals,and the same clinical manifestations may require different treatment options.The heterogeneity of cancer makes it impossible to assess tumors by relying on limited clinical indicators,which reflects the need to study cancer at the molecular level.In recent years,the rapid development of high-throughput sequencing technology and microarray technology has enabled researchers to systematically study cancer at the molecular level.For example,molecular biomarkers that predict the prognosis of cancer patients are found based on gene expression data.However,these prognostic genes often have poor generalization ability,and most of these genes are not oncogenes,but noise signals.Cancer is a system of multigene expression patterns and functional modules that are constantly changing,and it seems that the gene modules outperform the gene signatures in prognosis and molecular typing.Modules are not isolated,and there is also cross-talk among them.However,most studies have ignored the cross-talk,some important modules related to cancer might be overlooked.With advances in epigenetics research,the importance of DNA methylation abnormalities in cancer has gradually emerged.The genomic coverage of the DNA methylation microarray platform is increased and the cost is reduced,which has prompted us to study the pathogenesis of cancer at the methylation level.In our research,we collected DNA methylation data,gene expression data,and corresponding clinical data(including survival time and survival status)of breast invasive carcinoma(BRCA),skin cutaneous melanoma(SKCM)and uterine corpus endometrial carcinoma(UCEC)with abundant samples in The Cancer Genome Atlas(TCGA).First,we evaluated the stability in cancer prognosis of DNA methylation data and gene expression data for each of the three cancers and proved that DNA methylation data are more suitable for cancer prognosis research.Then,DNA methylation data were used to construct gene co-methylation networks for the three cancers using the rankbased method and to identify gene modules in three co-methylation networks.Next,we used the method of permutation testing to calculate the cross-talk between every two modules,thereby forming a module network,then we found the core gene modules in the module network by the K-shell method,and these core gene modules are used as features to study the prognosis and molecular typing of cancer.Finally,we found 2 core gene modules in BRCA,4 core gene modules in SKCM,and 2 core gene modules in UCEC.These core modules can significantly distinguish patients' prognosis.Then,these core modules as clustering features were used to classify three cancers by the Kmeans algorithm.The typing results were also significantly correlated with the prognosis of cancer patients.In addition,after analyzing the topology of the core module networks in three cancers,we identified DNA methylation prognostic biomarkers in three cancers.All these results demonstrate the effectiveness of our method in cancer prognosis and molecular typing.In summary,based on the core gene modules identified by the constructed DNA methylation module networks,we can distinguish not only the prognosis of cancer patients,but also use them for molecular typing of cancer.These results indicated that our method has important application value for the diagnosis of cancer,and may reveal potential carcinogenic mechanisms.