Fig. 1

Workflow of the study. Bioinformatics approach was composed of DNA methylation and gene expression analysis. DNA methylation data was obtained from experiment using Illumina Infinium HumanMethylation450k BeadChip (HM450) array from 381 tumor samples and 45 normal samples. Unsupervised clustering of tumor samples resulted in four clusters (CIMP-H, CIMP-L, Cluster 3 and Cluster 4). Probes of each cluster were compared to probes in normal samples group with Wilcoxon test to obtain differentially methylated probes. HM450 array contains 482,421 CpG sites, of which 190,920 CpG probes are located in promoter regions. The intersect of differentially methylated probes among all four methylation clusters and probes located in promoter region resulted in 3513 probes. Gene expression data was obtained from Illumina mRNAseq V2 experiment, which contains gene expression for 20,338 genes. From 381 tumor samples used in methylation analysis, 359 had gene expression data. For comparison of gene expression, 51 normal samples were used. Using gene expression data, tumor samples were divided into the same clusters as samples of methylation data and each cluster was compared to normal group using general log-linearized model to obtain differentially expressed genes. Intersect among all clusters in expression analysis gave 2422 differentially expressed genes. The bioinformatics approach resulted in 590 differentially methylated probes belonging to 198 differentially expressed genes, which exhibit hypermethylation/down-regulation or hypomethylation/up regulation. After literature mining, gene ontology, pathway analysis and protein-protein interactions we selected CEP55, FOXD3, FOXF2, GNAO1, GRIA4 and KCNA5 for further experimental validation. For experimental validation we used 115 samples in methylation experiment and 25 samples in gene expression experiment. After RNA/DNA isolation, DNA was bisulfite converted and used in methylation-sensitive high-resolution melt experiment (MS-HRM), and RNA was reverse transcribed to cDNA and used in quantitative real-time PCR (qPCR) experiment. Gene CEP55 was hypomethylated and up regulated, while FOXD3, FOXF2, GNAO1, GRIA4 and KCNA5 were hypermethylated and down-regulated, consistent with our bioinformatics analysis. Figure was prepared in Microsoft Power Point software. Legend: COAD, Colon adenocarcinoma; READ, rectum adenocarcinoma; logFC, logarithm of fold change