Abstract 7834: DNA methylation: A highly accurate biomarker for treatment monitoring - A retrospective case study using cfDNA methylation and machine learning in breast cancer patients.

Abstract DNA methylation plays a critical role in gene regulation and maintaining genomic stability. Abnormal DNA methylation patterns are common in breast cancer and relate to tumor growth and resistance to therapy. Liquid biopsies, especially analyzing cell-free DNA (cfDNA) in plasma, offer a minimally invasive way to monitor tumor-specific molecular changes in real time. This study evaluates the power of DNA methylation analysis to identify biomarkers for monitoring breast cancer treatment. Plasma samples from healthy individuals were used to first determine the reproducibility and robustness of the technology, while plasma from patients with breast cancer at different stages of disease was used for a biological validation study. After cfDNA extraction, all samples have been subjected to EM-seq library preparation followed by capture and next-generation sequencing. Sequenced reads were aligned to the human reference genome (hg38), and the DNA methylation levels were measured and filtered. The most differentially methylated CpGs were used as features for supervised machine learning models to differentiate between four disease states: progressive disease, partial remission, and complete remission, while some patients still showed residual abnormalities. The results of the analytical study demonstrate that the Hologic Diagenode Human Methylome procedure is highly reproducible across operators, input amounts, and technical replicates during the analytical phase. The biological validation demonstrates that, in patients undergoing various treatment regimens, the identified DNA methylation signature can predict their clinical state — ranging from progressive disease to partial and complete remission — with extremely high accuracy (92%). CpGs included in the identified signature were also relevant to the clinical context, as their associated genes were previously associated with cancer. These results underscore the potential of DNA methylation as a powerful molecular biomarker, not only for cancer diagnosis but also for treatment monitoring and potentially for minimal residual disease (MRD) detection. Citation Format: Jean-Valery Turatsinze, Andrea Blum, Adrien Godfroid, Ekaterina Gracheva, Matteo Tosolini. DNA methylation: A highly accurate biomarker for treatment monitoring - A retrospective case study using cfDNA methylation and machine learning in breast cancer patients abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7834.