pipette
ESEspañol

Enhanced cfDNA methylation profiling by detection and correction of jagged ends

Rui Wang, Emily G. Hamilton, Diego Almanza, Rene F. Bonilla, Takeshi Sugio, Angela Hui, Feng Tian, Nick A. Phillips, Chih Long Liu, Christopher H. Yoo, Stefan Alig, Cédric Rossi, Nicholas H. Juul, Emily Chen, Kevin J. Liu, Emily C. Shen, William T. Shi, Mari N. Olsen, Xiaoman Kang, Viswam S. Nair, Heather A. Wakelee, Joseph B. Shrager, Shuyu Shi, Tushar Desai, Mohammad Shahrokh Esfahani, Maximilian Diehn, Ash A. Alizadeh

Peer-reviewed journalReal-world use

In the authors' words

Analysis of cell-free DNA (cfDNA) methylation can enable noninvasive disease detection by leveraging tissue-specific epigenetic differences between malignant and nonmalignant cells. However, such analyses can be confounded by the repair of jagged DNA fragment ends during double-stranded sequencing library preparation, which introduces spurious methylation signals into cfDNA molecules. Here we present JEEPERS, a framework to detect and correct jagged-end-induced methylation errors while preserving duplex integrity and native methylation patterns. We showed high fidelity for measuring tumor burden and demonstrated allele-specific methylation differences at recurrent somatic mutations, revealing a link between adjacent CpG methylation and EGFR/TP53 driver and resistance mutations. We further demonstrated enhanced diagnostic performance across liquid biopsy applications, including tumor-type-specific methylation profiles, hemimethylation profiling, epigenetic-clock-based age estimation and cell type deconvolution. By faithfully correcting end-repair errors, JEEPERS provides a versatile framework for integrating mutations and methylation features in cfDNA at high resolution, enabling more precise molecular profiling and disease monitoring. The computational tool JEEPERS detects and corrects jagged-end-induced errors in cell-free DNA methylation analysis, demonstrating improved performance in multiple applications, such as cancer detection and monitoring.

Main resultThe abstract does not state a limitation.

Appeared: Saturday, October 31. Nature Methods. Peer-reviewed journal.

DOI: 10.1038/s41592-026-03278-5