Smarter, Cheaper, Faster AI Tool for Cancer Monitoring: Study Finds - Video

A team of scientists from the A*STAR Genome Institute of Singapore (A*STAR GIS) has developed an innovative artificial intelligence (AI)-based method called "Fragle" that simplifies and accelerates cancer tracking using blood tests. Their research, published in Nature Biomedical Engineering, demonstrates how Fragle can provide accurate, affordable, and faster monitoring of cancer by analyzing DNA fragment sizes in blood, without relying on expensive gene sequencing.

Tracking cancer using blood-based tests, known as circulating tumor DNA (ctDNA) analysis, is not new. However, existing methods depend heavily on detecting cancer-specific mutations, which can vary widely between individuals. This often results in inconsistent findings and high costs, limiting their usefulness for routine monitoring during treatment. Fragle addresses this challenge by focusing not on genetic mutations, but on the size patterns of DNA fragments in the blood—a unique feature of cancer DNA.

Fragle’s AI model can distinguish between healthy and cancerous DNA by analyzing fragment size patterns, using only a small amount of DNA. Fragle offers a faster and potentially more affordable way to monitor cancer through blood tests. “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,” explains Dr. Anders Skanderup, Senior Principal Scientist at A*STAR GIS and lead author of the study.

The method has been tested on hundreds of patient samples across various cancer types and is compatible with most hospital DNA profiling systems, allowing for rapid integration into clinical practice. Importantly, Fragle can detect minimal residual disease (MRD)—tiny traces of cancer left after treatment—helping catch relapse earlier than imaging scans.

Reference: Zhu, G., Rahman, C.R., Getty, V. et al. A deep-learning model for quantifying circulating tumour DNA from the density distribution of DNA-fragment lengths. Nat. Biomed. Eng 9, 307–319 (2025). https://doi.org/10.1038/s41551-025-01370-3