FragMIL: A Fragment-Token Multiple-Instance Learning Framework for Cancer Detection from Cell-Free DNA Fragmentomics
10 affiliations
- 1International Ph.D. Program in Medicine, College of Medicine, Taipei Medical University, Taipei 110, Taiwan
- 2Department of Pathology and Laboratory Medicine, Shin Kong Wu Ho-Su Memorial Hospital, Taipei 111, Taiwan
- 3AIBioMed Lab, Taipei Medical University, Taipei 110, Taiwan
- 4Department of Internal Medicine, Taipei Medical University Hospital, Taipei Medical University, Taipei 110, Taiwan
- 5Taipei Cancer Center, Taipei Medical University Hospital, Taipei Medical University, Taipei 110, Taiwan
- 6Department of Pathology, 103 Military Hospital, Hanoi 100000, Vietnam
- 7Institute of Molecular and Genomic Medicine, National Health Research Institutes, Miaoli 350, Taiwan
- 8In-Service Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University, Taipei 110, Taiwan
- 9Translational Imaging Research Center, Taipei Medical University Hospital, Taipei 110, Taiwan
- 10Graduate Institute of Clinical Medicine, College of Medicine, Taipei Medical University, Taipei 110, Taiwan
Abstract
Background/Objectives: Cell-free DNA (cfDNA) fragmentomics from low-coverage whole-genome sequencing (lcWGS) is a promising approach for non-invasive cancer detection. However, most existing methods summarize millions of cfDNA fragments into handcrafted sample-level features before classification, potentially discarding informative patterns carried by individual molecules. We aimed to develop and benchmark a framework that learns directly from fragment-level signals. Methods: We developed FragMIL, a fragment-token multiple-instance learning framework that represents each cfDNA molecule as a learnable token integrating end-motif, genomic-position, and fragment-length information, and aggregates thousands of fragment tokens into a patient-level cancer score. FragMIL was evaluated on 934 samples from three public lcWGS cohorts (pan-cancer, lung cancer, and hepatocellular carcinoma) using five repeated subject-level splits, a 288-configuration design-space benchmark, and comparison with four established fragmentomic approaches (coverage, DELFI, end motif, and fragment size). Results: FragMIL achieved the highest mean areas under the receiver operating characteristic curve (AUROCs) in every cohort, with 0.963 ± 0.009, 0.802 ± 0.009, and 0.910 ± 0.015 in the DELFI/Cristiano 2019, LUCAS/Mathios 2021, and Jiang 2015 cohorts, respectively. End-motif encoding provided the strongest standalone discriminative signal, whereas the optimal fragment pool was cohort-dependent, with open chromatin region fragments performing best for pan-cancer detection and tumor-enriched fragments for disease-specific cohorts. Performance was retained across cancer stages and cancer types. Conclusions: Preserving molecule-level fragmentomic information improves cancer detection, supporting fragment-token learning as a promising framework for next-generation cfDNA fragmentomics.
Related
- Weakly-Supervised Detection of Facial Motor Tics in Video Data Using Score-Driven Top-k Multiple Instance Learning
- Noninvasive Detection and Subtyping of Esophageal Cancer by Circulating Cell-Free mtDNA Fragmentomics: A Multi-Center Validation Study.
- Early and accurate detection of ovarian cancer by profiling cell-free mitochondrial DNA fragmentomics in multi-ethnic population.
- Multimodal Multiple Instance Learning for Fibroepithelial Tumor Diagnosis in Breast Ultrasound
- Multiple instance learning for differentiating neoplastic and non-neoplastic brain lesions using MR spectroscopy.
BibTeX
@article{bui2026fragmil,
title = {{FragMIL: A Fragment-Token Multiple-Instance Learning Framework for Cancer Detection from Cell-Free DNA Fragmentomics}},
author = {Tien-Dung Bui and Yung-Feng Lin and Mai-Hanh Nguyen and Shih-Feng Tsai and Nguyen-Quoc-Khanh Le and Teh-Ying Chou},
journal = {Cancers},
year = {2026},
month = oct,
doi = {10.3390/cancers18203389},
url = {https://www.mdpi.com/2072-6694/18/20/3389},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - FragMIL: A Fragment-Token Multiple-Instance Learning Framework for Cancer Detection from Cell-Free DNA Fragmentomics
AU - Tien-Dung Bui
AU - Yung-Feng Lin
AU - Mai-Hanh Nguyen
AU - Shih-Feng Tsai
AU - Nguyen-Quoc-Khanh Le
AU - Teh-Ying Chou
PY - 2026
DA - 2026/10/01
JO - Cancers
DO - 10.3390/cancers18203389
UR - https://www.mdpi.com/2072-6694/18/20/3389
AB - Background/Objectives: Cell-free DNA (cfDNA) fragmentomics from low-coverage whole-genome sequencing (lcWGS) is a promising approach for non-invasive cancer detection. However, most existing methods summarize millions of cfDNA fragments into handcrafted sample-level features before classification, potentially discarding informative patterns carried by individual molecules. We aimed to develop and benchmark a framework that learns directly from fragment-level signals. Methods: We developed FragMIL, a fragment-token multiple-instance learning framework that represents each cfDNA molecule as a learnable token integrating end-motif, genomic-position, and fragment-length information, and aggregates thousands of fragment tokens into a patient-level cancer score. FragMIL was evaluated on 934 samples from three public lcWGS cohorts (pan-cancer, lung cancer, and hepatocellular carcinoma) using five repeated subject-level splits, a 288-configuration design-space benchmark, and comparison with four established fragmentomic approaches (coverage, DELFI, end motif, and fragment size). Results: FragMIL achieved the highest mean areas under the receiver operating characteristic curve (AUROCs) in every cohort, with 0.963 ± 0.009, 0.802 ± 0.009, and 0.910 ± 0.015 in the DELFI/Cristiano 2019, LUCAS/Mathios 2021, and Jiang 2015 cohorts, respectively. End-motif encoding provided the strongest standalone discriminative signal, whereas the optimal fragment pool was cohort-dependent, with open chromatin region fragments performing best for pan-cancer detection and tumor-enriched fragments for disease-specific cohorts. Performance was retained across cancer stages and cancer types. Conclusions: Preserving molecule-level fragmentomic information improves cancer detection, supporting fragment-token learning as a promising framework for next-generation cfDNA fragmentomics.
SN - 2072-6694
ER -