Across two continents and two entirely different biological pathways to disease, a single AI-powered blood test has proven it can find liver cancer early — a quiet but consequential step in medicine's long effort to catch what kills before it becomes unstoppable. Researchers at Johns Hopkins have validated the DELFI platform using cell-free DNA analysis in populations from Guatemala and Romania, demonstrating that molecular fingerprints of cancer transcend geography, genetics, and cause. The test does not merely detect disease; it reads the biological story surrounding it, drawing signals from
AI blood test reliably detects liver cancer across diverse populations
These DNA fragments contain much more information than whether cancer is present
Mark
Why does it matter that the test works in both Guatemala and Romania? Couldn't you just validate it in one population?
Mimi
Because liver cancer is not one disease. In Romania it's mostly viral hepatitis and alcohol cirrhosis. In Guatemala it's metabolic disease and aflatoxin exposure. If the test only worked in one setting, you'd have no idea whether it was actually detecting the cancer or just the specific cause. This proves it detects something fundamental about the cancer itself, regardless of how it started.
Mark
The test looks at cell-free DNA fragments. Why is that better than the protein marker they use now—AFP?
Mimi
AFP is like looking for smoke from a fire. Cell-free DNA fragments are like examining the ash itself. They come from multiple sources—the tumor, the liver trying to repair itself, immune cells fighting back. You get a much more complete picture of what's happening in the body. AFP alone misses many early cancers.
Mark
You mentioned MethID, the technique that traces where fragments come from. How does that change things?
Mimi
It's the difference between knowing a crime happened and understanding the crime scene. You can now see which cells are contributing which signals. That lets you understand not just that cancer is present, but how the body is responding to it, which could eventually help predict which treatments will work.
Mark
The study mentions this could be a platform for detecting multiple diseases. How would that work?
Mimi
Instead of a separate blood test for each cancer, you'd run one test that analyzes the entire genome-wide pattern of DNA fragments. The artificial intelligence would learn to recognize the signature of lung cancer, liver cancer, pancreatic cancer—whatever you train it on. One blood draw, multiple answers.
Mark
What's the next step?
Mimi
Real-world validation. This was a carefully controlled study. Now they need to test it in actual screening programs where people come in without knowing whether they have cancer. That's where you find out if it really works in practice.
O Pulso
- Liver cancer is one of the world's deadliest and fastest-growing malignancies, yet current screening tools — ultrasound and a single protein marker — routinely miss tumors at the only stage when treatment can truly succeed.
- The DELFI blood test was put to a demanding cross-continental test: 377 patients in Guatemala and Romania whose cancers arose from metabolically, virally, and environmentally distinct causes — and it held up equally well across all of them.
- Rather than hunting for one marker, the test reads millions of DNA fragments circulating in the bloodstream, capturing signals from tumor cells, surrounding liver tissue, blood vessel walls, and immune cells — a richer biological portrait than any single biomarker can offer.
- Researchers even identified a mutation signature unique to aflatoxin exposure in Guatemalan patients that was absent in Romanian cohorts, yet the AI classifier remained accurate regardless — suggesting the platform adapts to population-specific biology without losing its universal footing.
- The vision is already expanding: a companion lung cancer screening test is now available in select U.S. health systems, and the team is building toward a single blood draw that could detect, locate, and characterize multiple diseases at once.
Across two continents and two entirely different biological pathways to disease, a single AI-powered blood test has proven it can find liver cancer early — a quiet but consequential step in medicine's long effort to catch what kills before it becomes unstoppable. Researchers at Johns Hopkins have validated the DELFI platform using cell-free DNA analysis in populations from Guatemala and Romania, demonstrating that molecular fingerprints of cancer transcend geography, genetics, and cause. The test does not merely detect disease; it reads the biological story surrounding it, drawing signals from tumor cells, immune responses, and damaged tissue alike. In a world where liver cancer is rising and early screening remains out of reach for many, this convergence of artificial intelligence and liquid biopsy points toward a more equitable diagnostic future.
A blood test that reads the molecular fingerprints of cancer has cleared a significant hurdle. Researchers at Johns Hopkins Kimmel Cancer Center validated their DELFI platform — which analyzes millions of cell-free DNA fragments circulating in the bloodstream — across 377 patients in Guatemala and Romania, two populations where liver cancer arises through strikingly different pathways. Romanian participants largely developed the disease through viral hepatitis and alcohol-related cirrhosis; Guatemalan participants through metabolic conditions and aflatoxin exposure from contaminated crops. Despite these biological differences, the AI-powered test identified liver cancer with consistent accuracy in both groups.
What distinguishes the approach is its depth of signal. Using a technique called MethID, researchers traced the origins of DNA fragments and found the test captures information not only from tumor cells but from surrounding liver tissue, blood vessel walls, and immune cells responding to the cancer. Combined with the protein marker AFP and basic clinical data, the test outperformed existing blood-based screening for both early and advanced disease. Co-senior author Zachariah Foda noted that these fragments reveal not just the presence of cancer but where it originates and how it evolves.
The team also found a mutation pattern linked to aflatoxin exposure that appeared exclusively in Guatemalan participants — yet the classifier remained effective across both cohorts, suggesting the platform can accommodate population-specific molecular variation without sacrificing universal accuracy. That adaptability carries real weight for global health equity, where liver cancer screening has long been limited or inaccessible.
The implications reach further still. Co-senior author Victor Velculescu envisions fragmentome analysis — the genome-wide study of cell-free DNA — as a foundation for detecting multiple diseases through a single blood draw. A companion lung cancer test, FirstLook Lung, is already available in certain U.S. health systems. Prospective clinical trials and multimodal refinements lie ahead, but the direction is clear: one vial of blood, read by artificial intelligence, may one day reveal not just whether disease is present, but where it came from and how best to treat it.
A blood test that reads the molecular fingerprints of cancer is working. Researchers at Johns Hopkins Kimmel Cancer Center have validated an artificial intelligence system that reliably detects liver cancer by analyzing millions of fragments of DNA circulating in the bloodstream—and the test performs equally well across populations with entirely different causes of disease.
The study, published July 31 in Cell Press Blue, tested the DELFI platform (DNA Evaluation of Fragments for Early Interception) on 377 people from Guatemala and Romania, two regions where liver cancer develops through strikingly different pathways. In Romania, most participants had contracted viral hepatitis or developed cirrhosis from alcohol. In Guatemala, the disease emerged primarily from metabolic causes—obesity, diabetes, and exposure to aflatoxin, a naturally occurring toxin that accumulates in contaminated crops. Despite these biological differences, the blood test identified liver cancer with consistent accuracy in both groups.
Liver cancer ranks among the world's deadliest malignancies, and its prevalence is climbing as metabolic disease spreads globally. Current screening relies on ultrasound imaging and a protein marker called alpha-fetoprotein, methods that frequently miss early-stage tumors when treatment options are broadest. The new test, when combined with AFP and basic clinical information like age and sex, detected both early and advanced cancers with greater sensitivity than existing blood tests alone.
What makes the approach powerful is not simply that it detects cancer—it reveals why. Using a technique called MethID, researchers traced where the DNA fragments originated. They discovered that the test captures signals not only from tumor cells but also from surrounding liver tissue, blood vessel walls, and immune cells mounting a response to the cancer. This multiplicity of signals creates a richer biological portrait than any single marker can provide. "These DNA fragments contain much more information than whether cancer is present," explained Zachariah Foda, an assistant professor at Johns Hopkins and co-senior author. "They tell us where these fragments originate and how they change during cancer development."
The team also identified molecular signatures unique to each population. Guatemalan participants showed a distinctive mutation pattern linked to aflatoxin exposure that did not appear in the Romanian cohort. Yet the overall classifier remained effective regardless of these regional differences, suggesting the test captures both universal features of liver cancer and population-specific molecular variations. This adaptability matters enormously for global health: a single platform could theoretically work across diverse genetic backgrounds and disease etiologies.
The implications extend beyond liver cancer. Victor Velculescu, the Cancer Genetics and Epigenetics Professor and co-senior author, noted that fragmentome analysis—the genome-wide study of cell-free DNA—could become a foundation for detecting multiple diseases through a single blood draw. The team has already moved forward with this vision: a companion test for lung cancer screening, called FirstLook Lung, is now available in certain U.S. health systems.
Future work will focus on prospective clinical validation—testing the approach in real-world screening settings—and refining multimodal strategies that combine fragmentome analysis with protein biomarkers and clinical risk factors. The researchers are building toward a future in which a single vial of blood, analyzed by artificial intelligence, could reveal not just whether disease is present but where it originated, how it is evolving, and what treatment might work best. For populations where liver cancer screening has been limited or inaccessible, that possibility represents a meaningful shift.
Citações Notáveis
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.— Victor Velculescu, Cancer Genetics and Epigenetics Professor, Johns Hopkins Kimmel Cancer Center
These DNA fragments contain much more information than whether cancer is present. They tell us where these fragments originate and how they change during cancer development, allowing us to better understand the biology of the disease.— Zachariah Foda, assistant professor of medicine, Johns Hopkins University School of Medicine
Quer a matéria completa?
Leia o original em News-Medical ↗
Technology