Humanmethylation27 array

Our proposed method was applied to two TCGA data sets: glioblastoma (GBM) and ovarian serous cystadenoma (OV). For both, we used the mutation, copy number, DNA methylation and expression data. We started our analysis with the Level 3 data downloaded from the TCGA website (https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix.htm). The GBM data set included 126 patients with mutation and copy number data, 235 patients with methylation and expression data, and 86 patients with mutation, copy number and expression data. The OV data set had 314 patients with mutation and copy number data, and 286 patients with mutation, copy number, methylation and expression data. For expression data for both GBM and OV, we used the U133A array which has 22,277 probes. For methylation data for GBM, we used the Illumina Golden Gate assay which has 1536 probes, and the HumanMethylation27 array which has 27,758 probes. We used the Golden Gate assay only when there were not enough samples with HumanMethylation27 data. For methylation data for OV, we used the HumanMethylation27 array which has 27,758 probes. In GBM, there were 673 mutations and 89,271 copy number alterations. In OV, there were 18,949 mutations and 13,859 copy number alterations.

Illumina-Infinium Human Methylation27 array was used for analysis of 24 samples including breast, colon, lung and endometrial epithelial tissues as well as white blood cells. Human sperm DNA and DNA treated with SssI CpG methyltransferase (NEB) were used as under-methylated and fully methylated DNA controls, respectively. Methylation scores (β values, 0–1) were generated for each of the 27,578 CpG loci (14,000 genes) on the array, based on the ratio of methylated to methylated+unmethylated signal-outputs. Partek® Genomics SuiteTM, version 6.5 was used for principal component analysis (PCA). β-average was calculated for each group of samples of the same origin, and selection of the most differentially methylated CpG loci was based on a β-average difference >0.2 and a P-value <0.05, after false discovery rate (FDR) correction, using analysis of variance (ANOVA). This part of the work was carried out at the genomics core service facility, the Rappaport research institute, Technion, Haifa, Israel.