Swedish researchers have quietly achieved something that could reshape how doctors track cancer treatment in real time: a blood test that can detect tumor DNA at levels four times lower than current clinical standards. Lotta Eriksson, a doctoral student at Chalmers University of Technology and the University of Gothenburg, helped develop BayesCNA, a new method that works with samples containing as little as 5% cancer DNA in the blood—a dramatic leap from the 15%-20% threshold required by today's techniques.
The breakthrough matters because cancer doesn't always announce itself loudly. When treatment works, tumor DNA in the bloodstream drops significantly, making it harder to monitor how a patient is responding. Current blood tests often fail in these quiet moments, when the very information doctors need most becomes hardest to extract. Traditional tissue biopsies require patients to undergo surgery—sometimes repeatedly—to glimpse how their cancer is evolving. Blood tests could change that calculus entirely, offering a window into tumor behavior at intervals of just a few weeks rather than months.
The team's innovation uses classical statistics to amplify extraordinarily weak signals buried in low-pass whole-genome sequencing data. Think of it as learning to hear a whisper in a crowded room. "You could compare it to skimming through a book rather than reading it properly," explains Eszter Lakatos, an assistant professor in the Department of Mathematical Sciences at Chalmers and the University of Gothenburg. "We get an overview of the DNA structure, but not a detailed picture." Yet by applying statistical algorithms to this overview, the researchers discovered they could extract far more information than anyone previously believed possible.
What surprised the team most was that classical statistics—the mathematical toolkit that feels almost quaint in an era of machine learning—outperformed cutting-edge AI methods at this particular task. "Nowadays, machine learning is used to solve a great many problems, and we tried those methods first," says Eriksson. "But, to our surprise, it turned out that classical statistics worked better in this case, which was particularly pleasing to us mathematicians and statisticians."
The implications ripple outward from there. BayesCNA can reveal details about a tumor's composition that were previously hidden in low-quality samples, painting a clearer picture of how a patient's cancer changes over time. Rather than waiting for the next surgery, oncologists could potentially adjust treatment strategies based on blood tests alone, tailoring therapy to match the tumor's actual composition between clinical sessions. For patients, that means fewer invasive procedures and more responsive care.
The research, published in Briefings in Bioinformatics, represents another step in the growing promise of liquid biopsies—blood tests that hunt for circulating tumor DNA. Multiple clinical trials worldwide are now exploring these techniques, but most have been constrained by the need for high cancer DNA concentrations. By lowering that threshold to 5%, the Swedish team has expanded the window of detection precisely when it matters most: when patients are responding to treatment and doctors need to stay one step ahead of the cancer.
The next phase is crucial. Researchers now aim to analyze what tumor composition information reveals about how individual patients respond to treatment, potentially unlocking the ability to predict who will benefit from which therapies. If successful, collaborations with clinical centers could soon follow—bringing these insights from the laboratory bench to the bedside.