Kirkegaard, J. B. | Search bioRxiv

bioRxiv2026

A Modular Framework for Automated Segmentation and Analysis of AFM Imaging of Chromatin Organization

Chromatin organization underlies essential genome functions, but its nanoscale organization remains challenging to capture and quantify with precision. Atomic force microscopy (AFM) offers direct structural readouts of DNA and chromatin, yet translating these rich images into reproducible biological metrics has been limited by the lack of standardized, scalable analysis tools. Here we present DNAsight, an automated analysis framework that integrates machine learning (ML)-based segmentation with modular, base-pair-calibrated quantification of DNA spatial organization, looping, nucleosome spacing, and protein clustering. Applied across diverse chromatin-associated proteins, DNAsight reveals protein-specific organizational signatures, including topology-dependent compaction by integration host factor (IHF), condition-dependent changes in loop-like DNA structures in cohesin-CTCF-precocious dissociation of sisters 5A (PDS5A) reactions, and promoter-driven multimerization of GAGA factor (GAF) clusters. The framework further enables direct extraction of nucleosome spacing distributions from raw AFM images, providing a label-free route to investigate chromatin fiber architecture. Together, these advances establish DNAsight as a generalizable and scalable approach for converting AFM measurements into quantitative insights into the physical principles of chromatin organization. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=79 SRC="FIGDIR/small/708946v2_ufig1.gif" ALT="Figure 1"> View larger version (31K): org.highwire.dtl.DTLVardef@1225883org.highwire.dtl.DTLVardef@1e1041dorg.highwire.dtl.DTLVardef@1d53d18org.highwire.dtl.DTLVardef@9e0a2a_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2023

Spontanously breaking of symmetry in overlapping cell instance segmentation using diffusion models

Instance segmentation is the task of assigning unique identifiers to individual objects in images. Solving this task requires breaking the inherent symmetry that semantically similar objects must result in distinct outputs. Deep learning algorithms bypass this break-of-symmetry by training specialized predictors or by utilizing intermediate label representations. However, many of these approaches break down when faced with overlapping labels that can appear, e.g., in biological cell layers. Here, we discuss the reason for this failure and offer a novel approach for instance segmentation based on diffusion models that breaks this symmetry spontaneously. Our method outputs pixel-level instance segmentations matching the performance of models such as cellpose on the cellpose fluorescent cell dataset while also permitting overlapping labels.

bioinformatics↗