BayesCNA Lowers Cancer DNA Detection Threshold to 5%, Enabling Non-Invasive Tumor Monitoring
Key Insights
Researchers at Chalmers University and University of Gothenburg developed BayesCNA (search), a statistical method that detects cancer (search) DNA in blood at levels as low as 5%, compared to the current 15–20% threshold.
The method uses classical statistics rather than machine learning to amplify weak signals from low-pass whole-genome sequencing, a cost-effective but low-resolution technique.
BayesCNA (search) could enable closer monitoring of tumor progression between treatment sessions, helping doctors tailor therapies based on tumor composition changes over time.
A team of researchers at Chalmers University of Technology (search) and the University of Gothenburg, Sweden, has developed a new analytical method that dramatically lowers the threshold for detecting and analyzing cancer (search) DNA in blood samples. The method, called BayesCNA (search), can reliably analyze samples containing as little as 5% cancer DNA—a significant improvement over the 15–20% required by current techniques.
The advance addresses a persistent challenge in liquid biopsy: when treatment is effective, the amount of circulating tumor DNA in the blood drops substantially, often falling below the detection limits of existing analytical methods. This makes it difficult to monitor how a patient's cancer (search) evolves over time.
"We wanted to develop a method that works particularly well in difficult cases where there is very little cancer (search) DNA in the blood and a lot of what we consider noise – that is, mainly healthy DNA," said Lotta Eriksson, a doctoral student in the Department of Mathematical Sciences at Chalmers and the University of Gothenburg. "Our results show that the new method performs better with samples involving low levels of cancer DNA, where the proportion is around 5 per cent. So, it works exactly as we had hoped."
A Statistical Approach to Signal Amplification
BayesCNA (search) was designed to analyze data from low-pass whole-genome sequencing, a technique that provides a broad overview of DNA structure at a fraction of the cost of more detailed methods. The trade-off has historically been poor data quality and limited information. The researchers liken the approach to skimming through a book rather than reading it in detail.
The new method employs a statistical algorithm to amplify the weak signals present in low-quality samples. Notably, the team initially explored machine learning approaches but found that classical statistical methods yielded superior results.
"Nowadays, machine learning is used to solve a great many problems, and we tried those methods first. But, to our surprise, it turned out that classical statistics worked better in this case, which was particularly pleasing to us mathematicians and statisticians," said Eriksson.
Clinical Implications for Treatment Monitoring
Current blood-based methods being tested in clinical trials are primarily used to determine whether cancer (search) can be detected at all. Obtaining a more detailed picture of tumor composition has been difficult due to high costs and poor sample quality. BayesCNA (search) can extract information previously hidden in low-quality samples, offering greater insight into a tumor's makeup.
"When the treatment is effective, the amount of cancer (search) DNA in the blood drops significantly. This makes it more difficult both to detect the cancer and to monitor how it changes. It is important to be able to analyze samples containing low levels of cancer DNA to gain a clearer picture of how a patient responds to treatment," said Eszter Lakatos, Assistant Professor in the Department of Mathematical Sciences at Chalmers and the University of Gothenburg.
At present, detailed information about tumor composition requires a tissue sample from the tumor itself. The ability to monitor tumor progression through blood tests could transform patient care by enabling more frequent, less invasive assessments.
"A patient may undergo surgery once or twice, whereas blood tests may be taken at intervals of just a few weeks during treatment. If we can obtain information about tumor changes from the samples, we can monitor developments much more closely and see what happens between treatment sessions. This can help doctors make more informed decisions, such as tailoring treatment to the tumor's composition," Lakatos added.
Next Steps Toward Clinical Application
The research team is now focused on analyzing the information BayesCNA (search) provides about tumor composition, with the goal of developing a further method for identifying hidden cancer (search) characteristics that influence treatment response.
"If we can demonstrate that this information is useful, we hope it will lead to more collaborations and wider adoption of our method within the research community. In the long term, I hope that the methods we develop can be used in clinical trials and, with any luck, make a difference to the care of cancer (search) patients," said Lakatos.
The development represents a meaningful step forward in the field of non-invasive cancer (search) monitoring, with the potential to enable more personalized and responsive treatment strategies for patients undergoing cancer therapy.