An AI-powered blood test for hepatocellular carcinoma detected cancer across 2 cohorts in Guatemala and Romania (n=377), despite markedly different underlying risk factors. In Cell Press Blue, the Johns Hopkins team reported the locked DELFI classifier detected 70% of cancers at 92% specificity, rising to 89% overall sensitivity and 79% in early-stage disease when combined with AFP, age and sex.

Why It Matters To Oncology

• The study suggests fragmentomics-based liquid biopsy can generalize across liver cancer populations shaped by different etiologies, including viral hepatitis, alcohol use, metabolic risk factors and aflatoxin exposure.

• For clinicians, the early-stage signal is notable: sensitivity was 54% in stage 0/A disease with DELFI alone across both cohorts, and improved when combined with AFP and clinical variables.

• MethID methylation analysis indicated the assay captures both tumor-derived DNA and host-response signals from liver, vascular and immune cells, which may inform future biomarker discovery.

The Financials

• No pricing, reimbursement or company revenue details were provided in the report.

• The commercial relevance is clear: a blood-based screening adjunct that improves on AFP could expand surveillance options in hepatocellular carcinoma, especially in heterogeneous at-risk populations.

What They're Saying

• Study author Victor Velculescu said the approach worked “with high performance across different patient populations while revealing the biological signals in the bloodstream that make this type of detection possible.”

• The authors also cautioned that the results require confirmation in larger prospective studies before broader clinical adoption.

What's Next

• The immediate next step is prospective validation in larger, real-world cohorts to test screening performance, especially in surveillance settings and earlier-stage disease.

• Researchers will likely further assess how DELFI plus AFP compares with current hepatocellular carcinoma surveillance workflows and whether fragmentomic and methylation signals can be translated into more robust multi-analyte assays.

• Given the metabolic-risk burden seen in Guatemala, future studies may also clarify performance in populations with diabetes mellitus (DM) and other metabolic liver disease drivers.