Identification Of Methylation Genes In Acute Myeloid Leukemia Based On High-throughput Data

ObjectiveBased on high-throughput microarray/sequencing data,bioinformatics analysis is used to mine gene methylation markers associated with AML,to analysis their possible action modes as well as mechanisms.Furthermore,the correlation between the important methylation genes and the patients’prognosis is explored,and the possible predictive value of those biomarkers is preliminarily evaluated.MethodsGEO and TCGA databases were retrieved systematically.The GSE58477methylated microarray data set,GSE35008 mRNA expression microarray data set in GEO,and AML-related methylated data set,RNA-seq data set in TCGA were screened as the analytical datasets according to our inclusion and exclusion criteria.1.For GEO data:⑴The impute package was used to fill in missing data of GSE58477.The wateRmelon package was used to standardize the data with the betaqn function.The quality control of data was made by drawing peak diagram and box diagram.The minfi package was used to analyze the differential methylation sites,as well as the bumphunter function was used to detect the differential methylation regions（DMR）between the AML samples and healthy control samples.Human hg19 reference genome was used for online annotation of the obtained differential methylation regions.GO analysis and KEGG pathway analysis for differential methylated genes were performed in the DAVID database.⑵The count data of GSE35008 mRNA expression microarray data set were read using R language and the affy package in Bioconductor environment,and the quality control was completed by arrayQualityMetrics package.The multi-array logarithmic robust algorithm（RMA）was used to treat the data background and to normalize it.The limma package was used to construct a linear model and a Bayesian test to identify differentially expressed genes,as well as performed an FDR adjustments on P-values.Volcano map and heat map of differential expression genes were drawn using gplots package.⑶The Venn diagram was drawn to identify the overlappd genes between differential methylation genes and differential expression genes.2.For TCGA data:⑴The Perl language was used to extract and integrate the count mRNA data matrix and the methylation data.⑵The edgeR software was used to normalize the gene expression data as well as the normalizeBetweenArrays function in the limma package was performed to normalize the methylation data.⑶For the overlapping genes screened in the first part,Spearman rank correlations were used to test the association between the gene methylation levels and their mRNA expression.⑷While 9 methylated genes were taken as the study variables,as well as age,gender,the first white blood cell counting were taken as covariates,Cox proportional hazard model was constructed to identify the methylated genes related with AML prognosis.⑸For the identified IFITM3 gene,Kaplan-Meier analysis was performed using the survival package,and the comparison between the two groups was performed by Log-rank test.Joint survival analysis of methylated levels and expression levels of IFITM3gene was performed by hash R package.⑹GSEA enrichment analysis of IFITM3,VAMP5 genes was performed by GSEA software.Results1.From GEO data:⑴Differential methylation analysis of GSE58477 screened a total of 43,357 differentially methylated sites,which showed the hypermethylation pattern in AML was more obvious than that of hypomethylation in the whole gene level.Further clustering analysis of the differentially methylated sites obtained a total of 562differentially methylated regions,revealed 647 differentially methylated genes in differentially methylated regions,which were involved in the regulation of multicellular biological processes(Bonferroni P=8.11×10^-08),positive regulation of RNA metabolism（Bonferroni P=0.0058）,biological processes of hematopoietic or lymphoid organ development（Bonferroni P=0.029）and biological pathways of HIF-1 signaling pathway（P=0.01）and osteoclast differentiation（P=0.017）.⑵differential expression analysis on GSE35008 mRNA data identified 540 differential expression genes（DEGs）.Differential expression and differential methylation analysis yielded 20 overlapping genes,including 7 genes with low methylation and high mRNA expression（MAP3K2、NAMPT、BCL7A、KDM2A、RAB11FIP1、TP53BP2、TRIB1）;6 genes with high methylation and low mRNA expression（CDC42BPA、GCNT2、GIMAP7、SCRN1、VAMP5、ZNF300）;5 genes with low methylation and low mRNA expression（ERCC4、IFITM3、MRPS23、STK33、STK335）;2 genes with high methylation and high mRNA expression（CDK12、MYH9）2.From TCGA data:⑴Correlation analysis of between the gene methylation level and their mRNA expression level in AML samples（n=125）showed there were 9genes in all 30 overlapped genes which the gene methylation levels were negatively related with their gene expression levels（all P<0.05）,which were BCL7A,CDC42BPA,GCNT2,GIMAP7,IFITM3,SCRN1,TRIB1,VAMP5 and ZNF300.⑵Cox analysis showed that IFITM3（adjustment OR=0.026,95%CI:0.002⁰.275）,VAMP5（adjustment OR=4.482,95%CI:1.277¹5.734）,age（adjustment OR=1.028,95%CI:1.010¹.046）and leukocyte result count（adjustment OR=1.008,95%CI:1.003¹.014）were related with AML prognosis.⑶GSEA enrichment analysis of single IFITM3 gene methylation showed that in the hypermethylation and hypomethylation groups,gene was mainly enrichment in cell adhesion molecules pathway（P=0.010,ES=0.482）and TCA cycle pathway（P=0.006,ES=-0.577）.The enrichment analysis of VAMP5 showed that in the hypermethylation and hypomethylation groups,gene was mainly enrichment in Type II diabetes pathway（P=0.016,ES=0.438）and TCA cycle pathway（P=0.002,ES=-0.592）.Conclusions1.Hyomethylation of BCL7A,TRIB1,IFITM3,hypermethylation of CDC42BPA,GCNT2,GIMAP7,SCRN1,VAMP5,ZNF300 was associated with AML susceptibility,which could be taken as Potential genetic methylation markers.2.Hypomethylation of IFITM3,hypermethylation of VAMP5 were associated with poor prognosis in AML patients.