Integrative Gastric Cancer DNA Methylation Analysis

We downloaded TCGA GC DNAm profiles (Illumina Human Methylation 450 BeadChip, Illumina Human Methylation 27 BeadChip), expression profiles, and corresponding clinical data through the UCSC Xena database (https://xena.ucsc.edu/) (Wang et al., 2019 (link)). The Illumina Human Methylation 450 BeadChip DNAm dataset contained two normal samples and 395 GC samples, while the Illumina Human Methylation 27 BeadChip DNAm dataset contained 25 normal samples and 48 GC samples. The expression profile dataset contained 32 normal samples and 372 GC samples. Table 1 lists the clinicopathological characteristics of the patients with GC. We downloaded the GC DNAm profile dataset GSE30601 from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) (Kurashige et al., 2016 (link)). The GSE30601 dataset was based on the GPL8490 platform (Illumina Human Methylation 27 BeadChip), containing 94 normal samples and 203 GC samples. The data from TCGA GC DNAm profiles were sorted and merged as the training dataset; the GEO GC DNAm profile dataset was used as the validation dataset. Because of the availability of public data in TCGA and GEO databases, this study did not require ethical approval or informed consent.