Johns Hopkins validated DELFI, an AI-powered blood test that analyzes cell-free DNA fragments to detect hepatocellular carcinoma across two high-risk cohorts in Romania and Guatemala (377 participants). Combining DELFI with AFP and basic demographics (age, sex) improved sensitivity for early and late disease. MethID analysis showed fragment patterns reflect tumor DNA and responses from liver, vascular, and immune cells. Next steps include prospective clinical trials and integrating protein biomarkers while current screening guidelines remain in place.
Johns Hopkins Validates AI Blood Test (DELFI) That Detects Liver Cancer Across Two High-Risk Populations
Photo Credit: iStock
An AI-driven blood test developed at the Johns Hopkins Kimmel Cancer Center has shown promise for earlier detection of liver cancer across two geographically and biologically distinct high-risk groups, according to a Johns Hopkins School of Medicine report shared by Medical Xpress.
What the Test Does
The assay, called DELFI (DNA Evaluation of Fragments for Early Interception), uses artificial intelligence to analyze millions of fragments of cell-free DNA circulating in the blood. Researchers report that DELFI can detect biological signals associated with liver cancer in the bloodstream, including tumor-derived DNA and changes in other cell types that respond to disease.
Study Design and Key Findings
The study, published July 31, analyzed blood samples from 377 participants recruited in Romania and Guatemala. Some participants had hepatocellular carcinoma (HCC), the most common form of liver cancer, while others did not.
These cohorts had very different underlying risk profiles: Romanian participants more often had liver disease linked to viral hepatitis or heavy alcohol use, whereas many Guatemalan participants had metabolic liver disease associated with obesity and diabetes and some had aflatoxin exposure. Despite these differences, DELFI performed well in both settings.
Investigators found that sensitivity for detecting both early- and late-stage tumors improved when DELFI results were combined with alpha-fetoprotein (AFP) levels plus simple demographic data (age and sex), compared with the fragmentome assay alone.
Why It Works: MethID Insights
Researchers applied a newer technique called MethID to map the biological origins of fragment patterns. MethID showed that fragmentomic signals reflect not only tumor DNA but also alterations in liver cells, blood vessels, and immune cells reacting to disease. Detecting signals from multiple tissue and cellular sources may help the test maintain performance across diverse patient populations.
Broader Implications and Next Steps
Because liver cancer is more treatable when found early, a more accurate, blood-based screening tool could expand access to surveillance and reduce reliance on imaging—especially for people at elevated risk from hepatitis, cirrhosis, heavy alcohol use, obesity, or diabetes.
The Johns Hopkins team is pursuing prospective clinical validation and developing assays that combine fragment analysis with protein biomarkers and established risk factors. Earlier work using the same platform also showed promise detecting liver fibrosis and cirrhosis, conditions that often precede HCC.
The researchers noted that a related DELFI-based screening test for lung cancer, FirstLook Lung, is already being used in some U.S. health systems through DELFI Diagnostics, underscoring the platform’s potential to support multiple noninvasive screening tests.
“This study demonstrates that the approach works with high performance across different patient populations while revealing the biological signals in the bloodstream that make this type of detection possible,” said Victor Velculescu, M.D., Ph.D., co-director of the cancer genetics and epigenetics program at Johns Hopkins Kimmel Cancer Center.
Clinical guidance: Existing screening recommendations remain unchanged for people with known liver disease or other risk factors until prospective validation is completed.
Help us improve.
