Scientists at the Johns Hopkins Kimmel Cancer Center have developed a liquid biopsy that identifies early‑stage cancers by measuring random variation in DNA methylation patterns. The approach, reported in Clinical Cancer Research, introduces the Epigenetic Instability Index (EII), which evaluates the degree of randomness—or “epigenetic instability”—in DNA methylation rather than its absolute levels. The study demonstrates that this method accurately distinguishes early‑stage lung and breast 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. “We immediately found that measuring DNA methylation variation performs better than just measuring DNA methylation by itself.”
Most liquid biopsies focus on detecting specific methylation changes at known genomic sites. While effective in defined groups, these tests often lose accuracy across diverse populations. To develop a more universal diagnostic tool, first author Sara‑Jayne Thursby analyzed publicly available datasets of 2,084 cancer methylation profiles to find 269 genomic regions—known as CpG islands—that captured most of the DNA methylation variability across several cancer types. “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,” Thursby said.
The team used these regions to train a machine learning model that could distinguish cancer signals from healthy ones and validated its performance using cross‑comparison techniques. The EII achieved strong sensitivity and specificity across multiple cancer types. In stage 1A lung adenocarcinoma, it detected cancer with 81% sensitivity at 95% specificity, while early‑stage breast cancer was identified with about 68% sensitivity at 95% specificity. The approach also showed promise in identifying colon, brain, pancreatic, and prostate cancers.
“Our hypothesis is that during the earliest stages of cancer development, methylation starts shifting,” Easwaran said. “We can try to pick those signals using these stochasticity metrics.” Co‑lead author Thomas Pisanic added, “By leveraging these metrics, we may be able to better identify and intercept tissues in the early stages of carcinogenesis.”
The team continues to refine the EII for broader clinical application, with plans to complement existing Johns Hopkins screening tools such as DELFI and assist in diagnostic decision‑making.
Publication details: Sara-Jayne Thursby, Zhicheng Jin, Jacob Blum, Andrei Gurau, Michaël Noë, Robert B. Scharpf, Victor E. Velculescu, Leslie Cope, Malcolm Brock, Stephen Baylin, Thomas Pisanic, Hariharan Easwaran; Epigenetic Instability Based Metrics in Cell-Free DNA for Early Cancer Detection. Clin Cancer Res 2026; https://doi.org/10.1158/1078-0432.CCR-25-3384
“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. “We immediately found that measuring DNA methylation variation performs better than just measuring DNA methylation by itself.”
Most liquid biopsies focus on detecting specific methylation changes at known genomic sites. While effective in defined groups, these tests often lose accuracy across diverse populations. To develop a more universal diagnostic tool, first author Sara‑Jayne Thursby analyzed publicly available datasets of 2,084 cancer methylation profiles to find 269 genomic regions—known as CpG islands—that captured most of the DNA methylation variability across several cancer types. “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,” Thursby said.
The team used these regions to train a machine learning model that could distinguish cancer signals from healthy ones and validated its performance using cross‑comparison techniques. The EII achieved strong sensitivity and specificity across multiple cancer types. In stage 1A lung adenocarcinoma, it detected cancer with 81% sensitivity at 95% specificity, while early‑stage breast cancer was identified with about 68% sensitivity at 95% specificity. The approach also showed promise in identifying colon, brain, pancreatic, and prostate cancers.
“Our hypothesis is that during the earliest stages of cancer development, methylation starts shifting,” Easwaran said. “We can try to pick those signals using these stochasticity metrics.” Co‑lead author Thomas Pisanic added, “By leveraging these metrics, we may be able to better identify and intercept tissues in the early stages of carcinogenesis.”
The team continues to refine the EII for broader clinical application, with plans to complement existing Johns Hopkins screening tools such as DELFI and assist in diagnostic decision‑making.
Publication details: Sara-Jayne Thursby, Zhicheng Jin, Jacob Blum, Andrei Gurau, Michaël Noë, Robert B. Scharpf, Victor E. Velculescu, Leslie Cope, Malcolm Brock, Stephen Baylin, Thomas Pisanic, Hariharan Easwaran; Epigenetic Instability Based Metrics in Cell-Free DNA for Early Cancer Detection. Clin Cancer Res 2026; https://doi.org/10.1158/1078-0432.CCR-25-3384