WAWABILITY July 11–12, 2025 Washington DC. Big ideas. Bold Progress. Global Impact. Powered by TDIforAccess.
WAWABILITY July 11–12, 2025 Washington DC. Big ideas. Bold Progress. Global Impact. Powered by TDIforAccess.

A-G.C.01: DifFracTion: Systematic normalization and differential interaction identification for cross-sample comparison of Hi-C datasets

Authors

Columbia University
Gamze Gursoy
Cambridge University

Keywords

Hi-C, Normalization, Differential Chromatin Interactions
[sponser-meet-now-chat][/sponser-meet-now-chat]
The three-dimensional (3D) organization of the genome plays a critical role in gene regulation, and disruptions to this architecture have been implicated in a wide range of diseases. Hi-C enables genome-wide measurement of chromatin interactions, motivating the development of methods for comparative analysis across biological conditions. However, most existing comparative approaches treat chromatin interactions as independent events, ignoring the strong distance-dependent decay of interaction frequencies, which limits their ability to detect differential chromatin interactions. Moreover, these methods fail when the Hi-C matrices being compared are sequenced at different depths. Here, we introduce DifFracTion, an algorithm that addresses these limitations through an iterative normalization approach that explicitly accounts for distance-dependent interaction frequency decay while correcting for sequencing depth imbalances, followed by a subsampling-based framework for estimating statistical significance. Under a strict null scenario with no biological differences, DifFracTion controls false positives across multiple resolutions, achieving zero false positive rate when sequencing-depth differences are modest and maintaining low FPR even under extreme imbalance. Additionally, DifFracTion accurately identifies true differential chromatin interactions across a wide range of fold-change magnitudes, achieving perfect accuracy and specificity, and exhibits a controllable trade-off between sensitivity and precision while outperforming both single- and multi-replicate approaches. Co-authors: Gamze Gursoy

Please login to see details