A context-aware analytical framework for DNA methylation biomarker discovery

Abstract

DNA methylation biomarker discovery commonly begins with tumor-associated differences at individual CpGs or genomic regions, yet the utility of a candidate depends on how its signal behaves across the biological backgrounds encountered in its intended application. Reference tissues, alternative tumor backgrounds and cohort composition can therefore shape whether a methylation change is informative. Here we present a gene-centered framework that integrates methylation contrast, intersample variability, context-specific methylation reference layers and sequence features to prioritize coordinate-defined genomic intervals. Applied to colorectal, hepatocellular and lung cancers, the framework revealed context-dependent shifts in candidate selection and regional metrics. Selected regions and a benchmark locus were then translated into native-DNA quantitative PCR assays that preserved discriminatory methylation signals in independent colorectal and liver tissue cohorts. Together, these results establish a context-aware route from genome-scale methylation data to experimentally addressable biomarker targets.