Deep-Learning-Enhanced TopoStats for the Automated Quantification of DNA and Complex Biomolecular Structures
- 1School of Chemical, Materials and Biological Engineering, University of Sheffield, Sheffield, UK .
- 2School of Medicine and Population Health, University of Sheffield, Sheffield, UK .
- 3Department of Computer Science, University of Sheffield, Sheffield, UK .
Abstract
Atomic force microscopy (AFM) enables nanometer-scale, label-free imaging of biomolecules and surfaces under near-native conditions, yet quantitative analysis of AFM data remains limited. This limitation largely arises from the absence of open, automated tools capable of addressing AFM-specific artifacts, data formats, and topographical outputs. Here, we present a substantial advancement to TopoStats, an open-source Python package for AFM image analysis, which introduces advanced analysis algorithms and deep learning capabilities to support more complex samples and richer molecular characterization. The new pipeline facilitates a shift from simple object segmentation and morphometric feature extraction toward automated tracing of molecules through complex crossings, enabling topological classification and curvature analysis. We demonstrate this by differentiating between plasmids with both different topology and sequence using quantitative descriptors, such as the number of crossings and median curvature, that distinguish the samples with statistical significance. Designed for accessibility and reproducibility, TopoStats adheres to the Findable, Accessible, Interoperable, and Reusable (FAIR) for Research Software principles and provides configurable workflows to allow the quantification of subtle structural changes within heterogeneous biological samples. This advances AFM from a fundamentally qualitative visualization technique toward a quantitative analytical tool, revealing properties not accessible with other structural biology techniques.
Related
- Propelling catalytic structures using active biomolecular condensates
- Automated Generation of Commensurate Magnetic Structures based on Spin Space Groups and Graph Theory
- Bayesian bilevel operator learning with low-rank adaptation for efficient uncertainty quantification of PDE inverse problems.
- Does System Size Matter in Uncertainty Quantification for Machine Learning Interatomic Potentials?
- Post-Selection-Free Quantum Automated Learning
BibTeX
@article{whittle2026deeplearningenhanced,
title = {{Deep-Learning-Enhanced TopoStats for the Automated Quantification of DNA and Complex Biomolecular Structures}},
author = {Sylvia Whittle and Tobias A. Firth and Max C. Gamill and Laura Wiggins and Neil Shephard and Toby Allwood and Daniel Brady and Thomas E. Catley and Alice L. B. Pyne},
journal = {Computational and Structural Biotechnology Journal},
year = {2026},
month = sep,
doi = {10.34133/csbj.0227},
url = {https://doi.org/10.34133/csbj.0227},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Deep-Learning-Enhanced TopoStats for the Automated Quantification of DNA and Complex Biomolecular Structures
AU - Sylvia Whittle
AU - Tobias A. Firth
AU - Max C. Gamill
AU - Laura Wiggins
AU - Neil Shephard
AU - Toby Allwood
AU - Daniel Brady
AU - Thomas E. Catley
AU - Alice L. B. Pyne
PY - 2026
DA - 2026/09/18
JO - Computational and Structural Biotechnology Journal
DO - 10.34133/csbj.0227
UR - https://doi.org/10.34133/csbj.0227
AB - Atomic force microscopy (AFM) enables nanometer-scale, label-free imaging of biomolecules and surfaces under near-native conditions, yet quantitative analysis of AFM data remains limited. This limitation largely arises from the absence of open, automated tools capable of addressing AFM-specific artifacts, data formats, and topographical outputs. Here, we present a substantial advancement to TopoStats, an open-source Python package for AFM image analysis, which introduces advanced analysis algorithms and deep learning capabilities to support more complex samples and richer molecular characterization. The new pipeline facilitates a shift from simple object segmentation and morphometric feature extraction toward automated tracing of molecules through complex crossings, enabling topological classification and curvature analysis. We demonstrate this by differentiating between plasmids with both different topology and sequence using quantitative descriptors, such as the number of crossings and median curvature, that distinguish the samples with statistical significance. Designed for accessibility and reproducibility, TopoStats adheres to the Findable, Accessible, Interoperable, and Reusable (FAIR) for Research Software principles and provides configurable workflows to allow the quantification of subtle structural changes within heterogeneous biological samples. This advances AFM from a fundamentally qualitative visualization technique toward a quantitative analytical tool, revealing properties not accessible with other structural biology techniques.
SN - 2001-0370
ER -