Related Experiment Video
Updated: Mar 13, 2026
11:11
Detection of Rare Mutations in CtDNA Using Next Generation Sequencing
Published on: August 24, 2017
17.4K
DeepFRAG: a method for cancer detection based on DNA fragmentomics and deep learning
1Laboratory Corporation of America Holdings (Labcorp), Burlington, NC 27215, United States.
Bioinformatics Advances
|March 12, 2026
Summary
DeepFRAG, a novel deep learning method, analyzes cell-free DNA fragment sizes for accurate and cost-effective cancer screening. This noninvasive approach shows high sensitivity and specificity, improving patient outcomes.
Area of Science:
- Oncology
- Bioinformatics
- Artificial Intelligence
Background:
- Liquid biopsy analyzing cell-free DNA (cfDNA) is crucial in cancer screening.
- Mutation-based diagnostics are sensitive but costly; cfDNA fragment size analysis offers a cost-effective alternative.
Purpose of the Study:
- To introduce DeepFRAG, a deep learning method for cancer detection using cfDNA fragment size distribution.
- To evaluate DeepFRAG's accuracy and effectiveness in diverse cancer types.
Main Methods:
- Deep learning analysis of cfDNA fragment size distribution profiles utilizing wavelet transform.
- Development of a novel data augmentation technique for whole genome sequencing (WGS) fragment size data.
- Validation on two independent cohorts (73 cancer patients, 80 healthy individuals).
Main Results:
- DeepFRAG achieved a median AUROC of 0.974.
- The method demonstrated high sensitivity (96.1%) and specificity (98.8%).
- Effective in detecting multiple major cancer types (breast, colorectal, pancreatic, lung, liver).
Conclusions:
- DeepFRAG offers a highly accurate, cost-effective, and robust noninvasive cancer detection method.
- This advancement expands options for early cancer screening and improved patient outcomes.
- Source code and data are publicly available for further research and application.
