A single drop of blood can now reveal which tissue is driving a disease

Liquid biopsies - blood tests that detect traces of disease without surgery or tissue samples - have been one of medicine's most promising but frustrating frontiers. The concept is elegant: tumor cells shed DNA fragments into the bloodstream, and if you can read those fragments, you can diagnose cancer from a simple blood draw. The problem is that current methods struggle to identify where those signals are coming from. Detecting that something is wrong is useful; knowing which organ or tissue is the source is far more clinically actionable.

Reading the epigenetic address labels

A team at Peking University, led by Professor He Aibin and Professor Jing Hongmei, has developed a platform that may solve this origin problem. Called cf-EpiTracing, the system captures detailed epigenetic fingerprints - chemical modifications to DNA that differ between cell types - from as little as 50 microliters of plasma. That is roughly a single drop of blood.

The approach works because different tissues leave distinct epigenetic marks on the cell-free chromatin circulating in blood. By reading these marks with multimodal analysis and machine learning algorithms, cf-EpiTracing can identify not just the presence of disease but the specific tissues driving it. The research was published in Nature in March 2026.

Colorectal cancer detection at 97.6% accuracy

In early diagnosis and screening for colorectal cancer, cf-EpiTracing delivered striking numbers. By integrating multiple epigenomic features from cell-free chromatin and applying machine learning, the platform reached an accuracy rate of 97.6% in training group samples and 92.2% in independent validation samples.

That validation figure is the more meaningful one. Training accuracy can be inflated by overfitting, but maintaining above 90% in a separate validation cohort suggests the signal is real and robust enough to have clinical potential.

Distinguishing lymphoma subtypes from blood

Beyond solid tumors, the platform revealed something unexpected in blood cancers. Patients with diffuse large B cell lymphoma showed stronger signals of CD34-positive cells in their plasma - a marker potentially reflecting bone marrow involvement and disease aggressiveness. This finding could offer new avenues for lymphoma subtyping and treatment stratification without requiring bone marrow biopsies.

The ability to distinguish cancer subtypes from a blood draw rather than invasive tissue sampling has been a long-standing goal in oncology. If cf-EpiTracing's subtyping capability holds up in larger studies, it could change how certain blood cancers are initially classified and monitored.

The gap between platform and clinic

It is important to be clear about where this technology stands. The results reported are from research cohorts, not from prospective clinical trials in real diagnostic settings. Moving from a published accuracy figure to a deployed clinical test involves regulatory approval, large-scale validation across diverse populations, and integration into existing diagnostic workflows - a process that typically takes years.

The sample sizes, while sufficient for initial demonstration, will need to be expanded substantially. Cancer screening tests require enormous validation studies to establish sensitivity and specificity in the general population, where the disease prevalence is low and false positives carry significant psychological and financial costs.

The researchers envision future directions that include integrating cf-EpiTracing with other cell-free modalities such as DNA methylation, mutations, and chromatin topology. This multi-omic approach could potentially increase precision further, but it also adds complexity and cost.

Why tissue-of-origin matters

The core advance here is not just detection sensitivity but tissue specificity. Many existing liquid biopsy approaches can detect circulating tumor DNA, but a positive signal without tissue-of-origin information leaves clinicians with a difficult question: where do we look next? A blood test that says "you likely have cancer in your colon" is fundamentally more useful than one that says "there are tumor fragments in your blood somewhere."

If the platform delivers on its promise across multiple cancer types and in larger populations, it could shift the liquid biopsy field from broad detection toward precise, tissue-specific diagnostics - which is where the real clinical value lies.