Abstract 3674: Predicting chromatin organization from cell-free DNA fragmentation patterns

Abstract Epigenetic regulation, including the three-dimensional (3D) architecture of the genome, plays a critical role in gene expression and cellular function. Aberrations in chromatin organization are implicated in various diseases, particularly cancer. While Hi-C is the gold standard for mapping chromosomal conformations, it has significant limitations: it requires fixed cells, limiting its capacity for longitudinal monitoring of dynamic chromatin changes. Cell-free DNA (cfDNA), which consists of short DNA fragments released into circulation through apoptosis and necrosis, offers a promising alternative. CfDNA fragmentation patterns are non-random and influenced by underlying epigenetic features, providing a window into chromatin organization in vivo. Here, we hypothesize that the correlation of cfDNA fragmentation patterns between two genomic regions can potentially reflect the 3D chromatin architecture. Since cfDNA fragments in vivo are released from multiple cell types, we generated in situ Hi-C and matched cfDNA whole-genome sequencing data from a xenograft mice model as the training dataset. We trained a contrastive deep learning model to infer Hi-C contact maps directly from cfDNA fragmentation data. The model demonstrated high accuracy in predicting chromatin compartments and topologically associating domains (TADs) from cfDNA fragmentation patterns, outperforming conventional machine learning approaches. These findings suggest that cfDNA fragmentation can enable the non-invasive and longitudinal monitoring of 3D genome aberrations in vivo. Citation Format: Ravi Bandaru, Zhihan Zhou, Hailu Fu, Weimin Wu, Haizi Zheng, Rashmi S. Hegde, Li Wang, Han Liu, Yaping Liu. Predicting chromatin organization from cell-free DNA fragmentation patterns [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3674.