Johns Hopkins researchers show that measuring DNA methylation variability can improve early cancer detection accuracy and strengthen liquid biopsy performance across diverse patient populations.
Researchers at Johns Hopkins Kimmel Cancer Center are advancing a new approach to liquid biopsy that could improve early cancer detection by focusing on variability in DNA methylation patterns—rather than absolute levels—offering a potentially more reliable biomarker across diverse patient populations.
Dark Daily’s sibling publication Today’s Clinical Lab reported that the liquid biopsy market is expected to increase by approximately 20% between 2022 and 2032, noting early cancer detection as a driver of the increase.
The method introduces a novel metric called the Epigenetic Instability Index (EII), designed to measure random variation, or “stochasticity,” in DNA methylation. In a proof-of-concept study published in Clinical Cancer Research, the approach demonstrated strong performance in distinguishing patients with early-stage cancers from healthy individuals.
“This is the first study where we are trying to really implement measuring that variation, or stochasticity, into a diagnostic tool,” said lead author Hariharan Easwaran, PhD. “We immediately found that measuring DNA methylation variation performs better than just measuring DNA methylation by itself.”
Model Targets Methylation Variability to Improve Multi-Cancer Detection
Traditional methylation-based liquid biopsies typically rely on detecting fixed changes at specific genomic sites. However, those tests are often developed using narrow patient cohorts and can struggle to generalize across broader populations. By contrast, the EII approach aims to capture a more universal biological signal tied to early tumor development.
To build the model, researchers analyzed more than 2,000 publicly available DNA methylation samples and identified 269 genomic regions (CpG islands) that capture the majority of methylation variability across cancer types.
“We identified specific genomic regions that tend to be the most variable in DNA methylation marks during cancer,” said first author Sara-Jayne Thursby, a postdoctoral researcher in Easwaran’s lab. “In cell-free DNA in the blood, that variability shouldn’t be high, but if it is, it is indicative of a developing cancerous phenotype.”
Using these regions, the team trained a machine learning model that demonstrated high accuracy across multiple cancers. In lung adenocarcinoma, the test detected stage 1A disease with 81% sensitivity at 95% specificity. For early-stage breast cancer, sensitivity reached approximately 68% at the same specificity level. The tool also showed potential utility in colon, pancreatic, brain, and prostate cancers.
Researchers say the findings support the idea that epigenetic instability may be an early hallmark of cancer progression.
“We hypothesize that early-stage tumors and precancerous lesions that exhibit high degrees of methylation variation… may be more resistant to intrinsic cancer-protective mechanisms and progress more rapidly,” said co-lead author Thomas Pisanic, PhD.
Looking ahead, the team plans to further validate the EII in larger clinical studies and position it as a complementary tool alongside existing screening methods. Easwaran noted that the test could serve as a “secondary triaging measure,” helping clinicians determine whether follow-up procedures—such as biopsies—are necessary after inconclusive or false-positive screening results.
For clinical laboratories, the approach signals a growing shift toward more nuanced, data-driven biomarkers that may improve early detection while reducing unnecessary procedures.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
—Janette Wider
