Background: DNA hypermethylation is a key event in oncogenesis and may act as a biomarker for the early detection of lung adenocarcinoma (LUAD). Here, we aimed to identify LUAD-specific methylation diagnostic biomarkers and explored potential mechanisms using data mining.
Methods: Using The Cancer Genome Atlas (TCGA) LUAD and GSE83842 datasets, we identified overlapping common differentially methylated positions (DMPs) with negative correlations between methylation and gene expression. Methylation profiles of the TCGA LUAD samples were compared with 185 blood samples and 370 lung squamous cell carcinoma (LUSC) samples to build a logistic regression model. Diagnosis performance was evaluated using an independent dataset.
Results: 160 genes were aberrantly methylated in LUAD since stage I; these genes were enriched in DNA-binding transcription factor activity, multiple embryonic development processes, and cell signaling. A diagnostic prediction model based on 10 CpG could distinguish LUAD from LUSC (area under the curve: 0.943). The derived model showed higher sensitivity and specificity than the two existing models. The homeobox A1 gene exhibited significantly higher methylation levels in LUAD than in 10 other cancers, showing potential as a LUAD-specific diagnostic biomarker.
Conclusions: Our findings provided insights into DNA methylation alterations in LUAD and established LUAD-specific diagnostic biomarkers.
Keywords: Diagnosis; Gene Expression Omnibus; Lung adenocarcinoma; Methylation; The Cancer Genome Atlas.