Liquid biopsy can detect that something is wrong without indicating where, which limits its clinical usefulness: a positive signal with no tissue of origin leaves the diagnostic problem largely unsolved. Researchers at Peking University, led from the College of Future Technology with the Department of Haematology at PKU Third Hospital, developed a platform called cf-EpiTracing that addresses this by capturing detailed epigenetic information from very small samples.
The sample requirement is as little as 50 microlitres of human plasma, roughly a single drop of blood. From that the platform extracts multimodal epigenomic features of cell-free chromatin, which carry signatures of the cell type the material came from, and applies machine learning to classify them. The approach therefore identifies the tissues driving a disease rather than only detecting circulating abnormality.
Performance was assessed across several conditions. In distinguishing colorectal cancer patients from healthy individuals the platform reached 97.6 per cent accuracy in the training group and 92.2 per cent in an independent validation group, and achieved a true detection rate of 77.27 per cent for colorectal adenoma, a precancerous lesion. In lymphoma, patients with diffuse large B-cell disease showed stronger CD34-positive cell signals in plasma, which may reflect bone marrow involvement and greater aggressiveness. The study was published in Nature.
Detecting precancerous adenoma is arguably the more consequential result, since colorectal screening gains most of its benefit from intervening before malignancy. The team plans to combine cf-EpiTracing with other cell-free markers, including DNA methylation, mutations and chromatin topology, in larger patient cohorts, which is the necessary step before accuracy figures from research cohorts can be trusted clinically.