DNA methylation signatures in
tumors could serve as reliable
biomarkers that are accessible in archival tissues for tracking the epigenetic dynamics shaped by both
cancer cells and the tumor microenvironment. However, given the ultrahigh dimensionality and noncollapsible nature of the data, it remains challenging to screen all CpG sites to identify the most promising marker panels. In this article, we introduce the concept of
tumor-based expression quantitative trait methylation (eQTM) for the prioritization and systematic mining of predictive
biomarkers. In
melanoma as a disease model, eQTM CpGs and genes represent new and efficient candidate targets to be investigated for both prognostic and immune status monitoring purposes. Three cis-eQTM CpGs (cg07786657, cg12446199, and cg00027570) were strongly associated with and can serve as surrogate
biomarkers for the
tumor immune cytolytic activity score (CYT). In addition, multiple eQTM genes could be further exploited for predicting immunoregulatory phenotypes. A targeted gene panel analysis identified one eQTM in TCF7 (cg25947408) as a novel candidate
biomarker for uncoupling overall T-cell differentiation and exhaustion status in a
tumor. The prognostic significance of this eQTM as an independent signature to CYT was validated by both The
Cancer Genome Atlas and Moffitt
melanoma cohort data. Overall, eQTMs represent a mechanistically distinct class of potential
biomarkers that can be used to predict patient prognosis and immune status.
SIGNIFICANCE: