mFrag: Integrating External DNA Methylation References Improves Lung Cancer Early Detection by Fragmentomics

mFrag improves lung cancer early detection from liquid biopsy fragmentomics by integrating external DNA methylation references, raising PR-AUC from 0.455 to 0.720 at 99% specificity.

• Why it matters: Fragmentomics can detect lung cancer non-invasively, but fragmentomic signals alone may be insufficient for early-stage accuracy, motivating integration with methylation context.
• What they did: mFrag combines mFrag-Weight (methylation-weighted fragmentomic features) and mFrag-Net (neural-network integration of fragmentomic and external methylation data), averaging their probabilities after training on 287 at-risk individuals.
• The result: On an independent case-control set (385 non-cancer controls; 46 cancers: 28 stage I, 12 stage II, 6 stage III or IV), mFrag increases sensitivity at 99% specificity from 0.043 to 0.478 and ROC-AUC from 0.871 to 0.908, and for stage I boosts PR-AUC 0.314→0.558 and ROC-AUC 0.855→0.894.

AI summary of the abstract · may contain errors · not medical advice · read the paper.