Pan-cancer Graph-based Cancer Detection Using the Cell-free DNA Methylome

Abstract

MotivationCell-free DNA methylation provides a minimally invasive signal for early cancer detection and tissue-of-origin prediction. Most methods represent methylation measurements as independent fixed-window features and therefore do not explicitly model relationships among genomic regions.

ResultsWe developed PANGEM (Pan-cancer Graph-based Cancer Detection Using the Cell-free DNA Methylome), a graph-learning framework that represents genomic bins as nodes and integrates CpG context, genomic proximity, and sample-specific methylation similarity in the graph topology. Across five repeated stratified train-test splits, PANGEM achieved the highest mean performance among evaluated methods, with an AUROC/AUPR of 0.997{+/-}0.003/1.000{+/-}0.000 for binary cancer detection and macro-AUROC/AUPR of 0.977{+/-}0.027/0.870{+/-}0.086 for multiclass tissue-of-origin prediction. In the independent INSPIRE cohort, 72 of 78 cancer cases (92.3%) exceeded the binary classification threshold, and PANGEM correctly classified 9 of 17 head and neck cancer cases (52.9%), the highest accuracy among evaluated methods. Subnetwork analysis further identified recurrent, graph-connected methylation patterns, including a 111-DMR subnetwork with increased methylation in cancer samples.

Availability and ImplementationSource code for this study is available at https://github.com/GenTensor/PANGEM. Contact: [email protected]

Supplementary informationSupplementary data are available at Bioinformatics online.

Noninvasive Detection and Subtyping of Esophageal Cancer by Circulating Cell-Free mtDNA Fragmentomics: A Multi-Center Validation Study.
Yue Liu, Wenying Yang, Yu Li +13
Advanced science (Weinheim, Baden-Wurttemberg, Germany)28 Oct 2026Peer-reviewedCC BYFull textSurgery & anaesthesiaEurope PMCPMC
Early and accurate detection of ovarian cancer by profiling cell-free mitochondrial DNA fragmentomics in multi-ethnic population.
Xinyan Guo, Zhiyun Gong, Ziyi Li +18
Genome medicine1 Jul 2026Peer-reviewedCC BY-NC-NDFull textObstetrics, gynaecology & paediatricsEurope PMCDOAJ
FragMIL: A Fragment-Token Multiple-Instance Learning Framework for Cancer Detection from Cell-Free DNA Fragmentomics
Tien-Dung Bui, Yung-Feng Lin, Mai-Hanh Nguyen +3
Cancers1 Oct 2026Peer-reviewedCC BY 4.0Full textOncologyCrossrefEurope PMC
Graph Neural Network Modelling of Pre-Diagnostic Serum Proteomics for Pancreatic Cancer Detection
Alexey Zaikin, Aleksandra Gentry-Maharaj, Sophia Apostolidou +5
Diagnostics (Basel, Switzerland)1 Oct 2026Peer-reviewedCC BYFull textGeneral medicineEurope PMC
Whole methylome changes in Barrett's metaplasia, dysplasia, and cancer
Y. J. Liew, T. Udumanne, K. Catalan +11
bioRxiv30 Oct 2026PreprintCC BYFull textOncologybioRxiv

Cite

BibTeX RIS

BibTeX

@misc{zhao2026pancancer,
  title = {{Pan-cancer Graph-based Cancer Detection Using the Cell-free DNA Methylome}},
  author = {L. Zhao and Y. Zeng and D. D. Abelman and W. Lin and P. Luo},
  howpublished = {medRxiv},
  year = {2026},
  month = aug,
  doi = {10.64898/2026.08.26.26361432},
  url = {https://www.medrxiv.org/content/10.64898/2026.08.26.26361432v1},
  note = {CC BY-NC-ND},
}

RIS

TY  - UNPB
TI  - Pan-cancer Graph-based Cancer Detection Using the Cell-free DNA Methylome
AU  - L. Zhao
AU  - Y. Zeng
AU  - D. D. Abelman
AU  - W. Lin
AU  - P. Luo
PY  - 2026
DA  - 2026/08/26
PB  - medRxiv
DO  - 10.64898/2026.08.26.26361432
UR  - https://www.medrxiv.org/content/10.64898/2026.08.26.26361432v1
AB  - MotivationCell-free DNA methylation provides a minimally invasive signal for early cancer detection and tissue-of-origin prediction. Most methods represent methylation measurements as independent fixed-window features and therefore do not explicitly model relationships among genomic regions.  ResultsWe developed PANGEM (Pan-cancer Graph-based Cancer Detection Using the Cell-free DNA Methylome), a graph-learning framework that represents genomic bins as nodes and integrates CpG context, genomic proximity, and sample-specific methylation similarity in the graph topology. Across five repeated stratified train-test splits, PANGEM achieved the highest mean performance among evaluated methods, with an AUROC/AUPR of 0.997{+/-}0.003/1.000{+/-}0.000 for binary cancer detection and macro-AUROC/AUPR of 0.977{+/-}0.027/0.870{+/-}0.086 for multiclass tissue-of-origin prediction. In the independent INSPIRE cohort, 72 of 78 cancer cases (92.3%) exceeded the binary classification threshold, and PANGEM correctly classified 9 of 17 head and neck cancer cases (52.9%), the highest accuracy among evaluated methods. Subnetwork analysis further identified recurrent, graph-connected methylation patterns, including a 111-DMR subnetwork with increased methylation in cancer samples.  Availability and ImplementationSource code for this study is available at https://github.com/GenTensor/PANGEM. Contact: [email protected]  Supplementary informationSupplementary data are available at Bioinformatics online.
ER  -

Abstract

Related

Cite