Advancements in Cancer Detection: A Statistical Breakthrough Enables Monitoring of Tumors Using Minimal Circulating DNA

Cancer detection and monitoring through blood tests, commonly referred to as liquid biopsies, have revolutionized the oncological landscape by providing less invasive alternatives to tissue biopsies. However, current technologies face significant limitations when it comes to analyzing samples containing low fractions of circulating tumor DNA (ctDNA), often struggling to detect and characterize cancer when ctDNA levels dip below 15 to 20 percent of total blood DNA. Researchers at Chalmers University of Technology and the University of Gothenburg, Sweden, have made a pioneering leap forward by developing an innovative analytical method capable of interpreting blood samples with as little as 5 percent tumor-derived DNA, potentially transforming cancer diagnostics and patient monitoring.

Traditionally, liquid biopsy methods have relied on the premise that a substantial proportion of the DNA in circulation originates from tumor cells, enabling the detection of genetic alterations that provide insight into tumor presence and composition. The difficulty arises when cancer DNA constitutes only a minor fraction amidst overwhelming healthy DNA, creating a noisy biochemical backdrop that hides crucial mutation signals. This complexity not only impedes early detection but also limits the ability to track tumor evolution and treatment response over time, especially when conventional treatment effectively reduces tumor burden and subsequently the ctDNA signal.

The newly developed approach, named BayesCNA, is a sophisticated statistical algorithm designed to distill meaningful information from low-pass whole-genome sequencing data of blood samples. Low-pass sequencing, which involves scanning the genome at a lower depth than traditional high-coverage sequencing, offers a cost-effective overview of genomic alterations but compromises detailed signal quality due to reduced data resolution. BayesCNA addresses this trade-off by leveraging classical Bayesian statistics to amplify subtle signals embedded within low-quality data, thus allowing the detection of copy number alterations—variations in the number of copies of specific DNA segments—that are hallmarks of many cancers.

This statistical ingenuity departs from the prevalent reliance on machine learning and artificial intelligence, which, although powerful, often require large datasets and high-quality inputs to perform optimally. Remarkably, the Chalmers team discovered that classical statistical modeling outperforms contemporary machine learning techniques in extracting reliable tumor-associated signals from samples heavily dominated by non-cancerous DNA. This finding demonstrates the enduring value of fundamental statistical principles in solving complex biomedical problems where data limitations challenge current computational paradigms.

The capability to accurately monitor tumor genetics from minimally invasive blood samples could revolutionize personalized oncology care. Currently, detailed tumor profiling necessitates obtaining tissue biopsies, which are invasive, sometimes risky, and infrequently performed throughout the course of treatment. In contrast, liquid biopsies can be administered frequently, offering a dynamic window into tumor biology as it responds and potentially adapts to therapy. Tracking changes in tumor genome composition via blood could inform clinicians if a treatment is diminishing tumor DNA levels or if resistance mechanisms are emerging, facilitating timely adjustments to therapeutic regimens.

Eszter Lakatos, Assistant Professor at Chalmers University of Technology and the University of Gothenburg, emphasizes the clinical implications: “When treatment is effective, the circulating cancer DNA plummets, making it difficult to detect the tumor’s signature in the blood. Our method excels in these challenging scenarios, unmasking tumor signals that would otherwise remain undetectable.” This enhanced sensitivity offers profound opportunities for early detection of relapse and refinement of treatment strategies based closely on evolving tumor profiles.

The BayesCNA method provides an analytical breakthrough by focusing on the sensitive detection of copy number alterations in complex samples, which are critical markers of tumorigenesis and disease progression. By decoding these genomic aberrations from skimmed sequencing data, researchers and clinicians gain access to deeper insights without incurring prohibitive costs or demanding ultra-high sequencing depths. This balance promises practical integration into clinical workflows and broader accessibility across healthcare systems.

Underlying the success of this method is the application of Bayesian inference, a statistical framework that updates the probability for a hypothesis as more evidence or information becomes available. In the context of low-pass genome sequencing, Bayesian strategies allow the algorithm to incorporate prior knowledge about tumor biology and sequencing noise, refining its predictions even amidst uncertainty. This contrasts with many machine learning models that may struggle to incorporate such domain-specific priors or interpret overtly noisy datasets.

Lotta Eriksson, doctoral student and study co-author, reflects on the methodological choice: “Initially, we experimented with various machine learning tools, expecting them to be superior. It was surprising and gratifying that classical statistics delivered more robust and interpretable results. This not only validates our mathematical approach but highlights the importance of matching problem-solving techniques to the nature of biomedical data.”

Looking ahead, the research team aims to expand their analytical framework to discern additional hidden features of tumors that influence patient responses to various treatments. The ultimate goal is to translate these quantitative insights into actionable clinical decision support, enabling oncologists to tailor therapies with unprecedented precision based on real-time tumor genomics.

The promise of integrating BayesCNA into clinical trials is profound. Widespread adoption could accelerate the standardization of blood-based tumor monitoring, reducing dependence on invasive biopsies and improving patient outcomes through personalized care pathways. Frequent sampling and sensitive analysis provide opportunities to preemptively identify therapeutic resistance, adapt dosing strategies, or employ combinatorial treatments to forestall disease progression.

In the context of healthcare economics, the reduced sequencing depth required by BayesCNA translates into lower costs, which is a critical consideration for the scalability of genomic diagnostics. The ability to leverage low-pass whole-genome sequencing without sacrificing analytical power democratizes access to cutting-edge cancer monitoring tools, especially in resource-constrained settings.

This advancement also underscores the interdisciplinary collaboration between mathematical sciences and biomedical research, highlighting how computational innovation can drive substantial improvements in clinical applications. By harnessing statistical expertise and leveraging domain knowledge on tumor biology, the Chalmers and Gothenburg teams have charted a course toward a new paradigm in oncology diagnostics.

Ultimately, BayesCNA represents a transformative leap in liquid biopsy technology’s capability, potentially heralding a future where cancer is continuously monitored with high resolution through routine blood tests. Such progress holds the promise of not only improving survival rates through timely interventions but also enhancing the quality of life for patients by minimizing invasive procedures and personalizing therapeutic strategies.

Subject of Research: People

Article Title: Sensitive detection of copy number alterations in low-pass liquid biopsy sequencing data

News Publication Date: 16-Mar-2026

Web References: https://doi.org/10.1093/bib/bbag111

References:
Eriksson, L., & Lakatos, E. (2026). Sensitive detection of copy number alterations in low-pass liquid biopsy sequencing data. Briefings in Bioinformatics. https://doi.org/10.1093/bib/bbag111

Image Credits: Chalmers University of Technology | Marco Nikic

Keywords: Liquid biopsy, circulating tumor DNA, copy number alterations, low-pass whole-genome sequencing, Bayesian statistics, cancer monitoring, tumor evolution, personalized cancer treatment, bioinformatics, statistical modeling

Tags: advanced cancer biomarker analysisblood-based tumor monitoringCancer diagnostics innovationChalmers University cancer researchcirculating tumor DNA analysisearly cancer detection methodsgenetic mutation detection in bloodliquid biopsy cancer detectionlow fraction ctDNA detectionminimally invasive cancer monitoringnon-invasive oncology testingtumor DNA statistical breakthrough