Tumor-agnostic multimodal cell-free DNA analysis improves detection and enables longitudinal disease monitoring in esophageal adenocarcinoma
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Esophageal adenocarcinoma (EAC) is an aggressive malignancy for which liquid biopsy–based clinical monitoring using cell-free DNA (cfDNA) remains challenging due to low tumor cfDNA shedding, few recurrent mutations, and limited access to tissue required for tumor-informed assays. Moreover, genome-wide cfDNA analysis remains largely unexplored in prospective longitudinal EAC cohorts. Methods Here, we developed a tumor-agnostic multimodal cfDNA framework integrating genomic and fragmentomic features. We applied this framework to 137 prospectively collected longitudinal plasma samples from 49 patients using deep targeted sequencing, personalized digital PCR, and shallow whole-genome sequencing (sWGS). Using sWGS-derived cfDNA features, disease-status and 6-month mortality classifiers were developed using elastic-net logistic regression and evaluated in held-out test cohorts. Results At baseline, targeted sequencing detected tumor cfDNA in only 52% of patients, while integrating targeted mutation analyses with genome-wide copy-number alterations and fragmentomic features increased detection to 85%. Longitudinal profiling detected tumor cfDNA in all patients with plasma available at clinical relapse. In held-out test cohorts, the classifiers achieved AUCs of 0.82 for disease status and 0.81 for 6-month mortality risk. Conclusions Together, these findings establish that integrating complementary mutation, copy-number, and fragmentomic signals improves cfDNA detection in EAC, while machine-learning analysis of features from a single low-cost, tumor-agnostic sWGS assay provides disease and prognostic information. These results support sWGS as a scalable strategy for longitudinal liquid-biopsy assessment in EAC, with potential applicability to other low-shedding cancers.
Source: Research Square
Authors: Alexandra Bartolomucci, Lisa-Monique Edward, Sarah Tadhg Ferrier, Jeffrey P. Bruce, Dor D. Abelman, Monyse Nobrega, Kwang Bo Joung, Arnavaz Danesh, Stephenie D. Prokopec, Luis Santos Castro, Abirami P. Sharma, Rawan Sakalla, Wotan Zeng, Carolina Martini, Kyle Dickinson, Nicholas Bertos, Lorenzo Ferri, Jonathan Cools-Lartigue, Trevor J. Pugh, Julia V. Burnier
Language: English
Subfield: Cancer Research
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7221762005