Reviewed by Lexie CornerJun 16 2025
Researchers at the ASTAR Genome Institute of Singapore (ASTAR GIS) have developed an artificial intelligence (AI)-based method called “Fragle” to help track cancer through blood tests. The findings were published in Nature Biomedical Engineering in March 2025.
The method uses a small blood sample and analyzes the size of DNA fragments to identify patterns that differentiate cancer DNA from healthy DNA. This allows for more precise and frequent monitoring of cancer treatment response.
Current methods for detecting circulating tumor DNA (ctDNA) often rely on extensive and expensive sequencing to identify common cancer mutations. Since mutations vary between patients, results can be inconsistent, making it harder to evaluate treatment response using blood tests.
Fragle uses AI to measure DNA fragment size, which differs between cancerous and healthy DNA. The model can detect these differences using minimal DNA input. This makes the process faster and less costly than traditional methods.
The technique has shown consistent accuracy across various cancer types and hundreds of patient samples. It is also compatible with most DNA profiling methods used in hospitals and by commercial testing providers
Key Benefits of Fragle:
Faster and More Affordable:
Fragle offers a quicker and potentially more cost-effective way to monitor cancer using blood tests, requiring only a small amount of DNA. While conventional commercial tests can cost over SGD $1,000, Fragle is expected to cost under SGD $50.
Compatible With Existing Tools:
Fragle works with commonly used DNA profiling technologies in hospitals and commercial labs, making it easy to integrate into current diagnostic workflows.
Early Detection of Relapse:
Fragle can detect very small amounts of cancer remaining after treatment—known as minimal residual disease (MRD)—helping doctors identify potential relapse at an early stage.
Just as scientists tracked COVID-19 outbreaks by detecting viral particles in wastewater, Fragle analyzes DNA fragments in blood to monitor cancer treatment response and detect relapse early. While existing methods have their strengths, they are often complex and expensive. We wanted to develop a simpler, more affordable, and accessible approach, one that could support accurate monitoring without adding burden to clinical workflows.
Dr Anders Skanderup, Study Lead Author and Senior Principal Scientist, A*STAR Genome Institute of Singapore
The team is currently working to improve Fragle’s sensitivity to detect even smaller amounts of cancer DNA. This is important for identifying early signs of disease recurrence in cancer patients.
They are also collaborating with the National Cancer Centre Singapore (NCCS) to explore possible clinical applications. The team plans to study how Fragle could be implemented in local hospitals to support cancer care.
We are excited to initiate studies on how methods such as Fragle can detect disease relapse earlier in local lung cancer patients.
Daniel Tan, Study Co-Author and Associate Professor, Division of Medical Oncology, National Cancer Centre Singapore
In an ongoing study involving over 100 clinical trial participants, the GIS-NCCS team is using Fragle to monitor ctDNA levels every two months during treatment. The goal is to detect signs of relapse before they appear on standard imaging scans.
The researchers are also examining whether early changes in ctDNA levels can help predict how well a patient will respond to therapy. The study aims to evaluate the potential of using ctDNA testing as part of routine cancer monitoring during treatment..
We are very excited about the potential Fragle brings, to help our healthcare professionals detect and track cancer more accurately and monitor treatments more effectively, leading to better cancer care for patients. It is our hope that our genomic research can be translated to benefit population health not only in Singapore, but worldwide.
Dr Wan Yue, Executive Director, A*STAR Genome Institute of Singapore
Journal Reference:
Zhu, G., et al. (2025) A deep-learning model for quantifying circulating tumour DNA from the density distribution of DNA-fragment lengths. Nature Biomedical Engineering. doi.org/10.1038/s41551-025-01370-3.