ChromaFactor: Deconvolution of single-molecule chromatin organization with non-negative matrix factorization
Laura M. Gunsalus, Michael J. Keiser, Katherine S. Pollard
Abstract
The investigation of chromatin organization in single cells holds great promise for identifying causal relationships between genome structure and function. However, analysis of single-molecule data is hampered by extreme yet inherent heterogeneity, making it challenging to determine the contributions of individual chromatin fibers to bulk trends. To address this challenge, we propose ChromaFactor, a novel computational approach based on non-negative matrix factorization that deconvolves single-molecule chromatin organization datasets into their most salient primary components.
Introduction
Chromatin is intrinsically dynamic, and its behavior across time restricts and permits the precise regulatory landscape controlling gene expression [1,2]. Recent single-cell technologies such as single-cell Hi-C [3,4] and chromatin microscopy techniques [5–9] now offer unique insight into genome folding, allowing us to directly observe chromatin folding as well as functional readouts in individual cells to disentangle their mechanistic relationship.
Methods
Datasets and processing
Mateo et al. dataset.
The Mateo et al. microscopy dataset contains 3D genomic coordinates and transcriptional activity for single molecules spanning the Drosophila Bithorax complex (BX-C) locus [5], which can be found at the following repository: https://zenodo.org/records/4741214.
Results
NMF to decompose single-cell 3D genome conformation datasets
We were motivated to develop ChromaFactor by the disconnect between meaningful signal observed in bulk cell populations and the extreme heterogeneity of single-molecule examples. One such dataset, Mateo et al. 2019 [5], profiles local chromatin conformation at a single locus - the bithorax complex (BX-C) - in Drosophila embryos and additionally includes matched nascent transcription in the same cells (n = 16,320). To discover how cell populations vary, we often take the difference between the average contact maps under two conditions. We observed a pronounced boundary in cells within the 30 kb region actively transcribing the Abd-A gene, as compared to non-transcribing cells (Fig 1a).
Discussion
This study has demonstrated the utility of ChromaFactor, a novel application leveraging Non-negative Matrix Factorization (NMF) for dissecting single-cell chromatin conformation datasets. Our method uncovers nuanced layers of genome conformation dynamics and their correlation with transcriptional states, which would otherwise be obscured in bulk analyses. Correlations between template patterns and active transcription suggest that these templates are not merely reflections of cellular heterogeneity, but could be mechanistically linked to transcriptional regulation.
Acknowledgments
We thank Chris Olah for the original idea to apply NMF to single-cell 3D genome datasets and Kangway Chuang for project guidance, including the idea to train a random forest to test component significance. We additionally thank Archit Verma, Amanda Everitt, Katie Gjoni, Shuzhen Kuang, and other members of the Pollard and Keiser labs for helpful discussion and manuscript feedback.
Citation: Gunsalus LM, Keiser MJ, Pollard KS (2025) ChromaFactor: Deconvolution of single-molecule chromatin organization with non-negative matrix factorization. PLoS Comput Biol 21(2): e1012841. https://doi.org/10.1371/journal.pcbi.1012841
Editor: Jie Liu,, University of Michigan, UNITED STATES OF AMERICA
Received: May 19, 2024; Accepted: February 2, 2025; Published: February 18, 2025
Copyright: © 2025 Gunsalus et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All code used for the method, running experiments and creating figures is available on a GitHub repository at https://github.com/lgunsalus/ChromaFactor.
Funding: This work was supported by the National Institutes of Health 4D Nucleome Project (#U01HL157989 to KSP), the Chan Zuckerberg Initiative DAF (an advised fund of the Silicon Valley Community Foundation) (#2018-191905 to MJK), and the University of California San Francisco (Achievement Rewards for College Scientists Scholarship to LMG). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. KSP and LMG received salary support from the National Institutes of Health. LMG received salary support from the University of California San Francisco. MJK received salary support from the Chan Zuckerberg Initiative.
Competing interests: The authors have declared that no competing interests exist.
