• Citer

From Tumor Informed to Fragmentomics: Expanding the Toolkit for ctDNA Detection and Quantification approaches in Breast Cancer

par Stanciu, Maria-Alexandra
Président du jury Verhoeven, Caroline
Promoteur Ignatiadis, Michail
Publication Non publié, 2026-03-05

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

CtDNA, released into the bloodstream by cancer cells, has emerged as a powerful biomarker for detecting and monitoring cancer through liquid biopsies. However, its detection and interpretation remain challenging, particularly in early-stage and low-shedding tumors such as hormone receptor–positive, HER2- negative (HR+/HER2–) breast cancer. This thesis explores complementary strategies to improve ctDNA detection and quantification for minimal residual disease (MRD) detection, treatment monitoring, and risk stratification in breast cancer. Two distinct approaches were pursued: (i) personalized ctDNA detection using the RaDaR assay in patients with early breast cancer, and (ii) quantitative ctDNA estimation in metastatic breast cancer women based on cell free DNA (cfDNA) fragmentation patterns using ALFAssay, a deep learning model.We assessed the ctDNA detection using a commercially developed assay to patients enrolled in the NeoRHEA trial, investigating neoadjuvant endocrine therapy combined with CDK4/6 inhibition in HR+/HER2– early breast cancer. ctDNA was detectable in more than half of patients before treatment and was associated with aggressive tumor characteristics such as high grade, increased proliferation, and unifocal growth. Importantly, persistence of ctDNA during or after therapy was linked to poor pathological response, suggesting that ctDNA could serve as a sensitive, non-invasive marker of residual disease and therapeutic resistance. This study adds to the growing body of evidence supporting the clinical validityof ctDNA analysis in the neoadjuvant setting, particularly for hormone receptor–positive breast cancer, where molecular monitoring tools remain challenging.Complementary to mutation based assays, a novel computational model, ALFAssay, was developed to quantify tumor burden based on cfDNA fragmentation features obtained from shallow whole-genome sequencing. Unlike conventional mutation- or copy-number-based methods, ALFAssay captures fragmentation patterns reflecting tumor-derived cfDNA. Across nearly 900 plasma samples from multiple breast cancer and healthy cohorts, ALFAssay effectively distinguished cancer from healthy plasma, showed high concordance with existing algorithms (ichorCNA and Fragle), and provided prognostic information in patients with HR+/HER2– metastatic breast cancer enrolled in the Pearl study, which assessed ctDNA and PET imaging as tools for monitoring endocrine therapy efficacy. Patients with high ctDNA fractions or persistent ctDNA during treatment had significantly shorter progression-free survival (PFS), underscoring the prognostic relevance of ctDNA kinetics. Quantifying ctDNA remains challenging and a few tools are available to accurately estimate tumor-derived cfDNA fractions. In this context, ALFAssay provides an important addition to the liquid biopsy toolkit by leveraging cfDNA fragmentation features to infer tumor fraction independently of genomic alterations. This study demonstrates ALFAssay model potential to complement existing approaches such as targeted sequencing and ichorCNA, which rely on mutation or copy-number detection, as well as Fragle, which also exploits fragmentation information but captures distinct biological signals. By expanding the analytical framework for cfDNA quantification, ALFAssay contributes to a more comprehensive and biologically informed understanding of ctDNA dynamics.This work advances the understanding and application of ctDNA analysis across different breast cancer settings. By evaluating a commercially developed assay in early HR+/HER2– disease, we provided valuable evidence of its performance in a context where ctDNA detection remains technically and biologically challenging. In parallel, we developed ALFAssay, a fragmentation-based computational model that quantifies ctDNA in the metastatic setting, and which shows complementarity with existing mutation- and copy-number–based approaches. Together, these studies contribute to refining cfDNA-based monitoring strategies and highlight the potential of integrating diverse molecular signals to improve disease assessment and treatment evaluation in breast cancer.