AI tool untangles DNA knots to help predict health impacts

DNA bends, loops, and crosses itself in cramped cells, and those ‘knots’ can help or hinder life. Scientists have long taught the neat ladder-like twist, but spent years wrestling with the messier truth.

Researchers have now developed a rapid method to visualize DNA crossings and determine, at the level of single molecules, which strand passes over or under the other.

The method pairs high-resolution imaging with smart software and quickly delivers answers that used to take a lot of time. The research was led by Alice Pyne, a professor of biophysics at the University of Sheffield.

Mapping DNA’s hidden knots

The group imaged DNA using atomic force microscopy (AFM) and then traced each molecule’s path with a deep-learning pipeline that identifies every crossing and labels it as over or under.

The team demonstrated that they can recover the topology, the length, and the local shape of single DNA circles and their tangles with sub-molecular detail.

Unlike light-based microscopes, atomic force microscopy feels the surface with a nanoscale probe, which makes it well suited to single molecule work in fluid.

The role of atomic force microscopy

A previous study showed that atomic force microscopy can map DNA, without staining, under near-cellular conditions.

The Sheffield pipeline advances this by measuring the height profile at each strand crossing and applying a full-width-at-half-maximum comparison to identify which DNA strand passes over. This method increases accuracy, especially when crossings occur in close proximity.

The software then outputs a topological class for the whole molecule, so researchers do not have to infer knot types by gel position or by eye.

DNA knots and disease

Cells rely on DNA topology to manage access to genes and to keep the genome intact. When the balance tips, damage builds up and repair systems struggle.

A comprehensive review explains how each human topoisomerase family member cuts and rejoins DNA to manage supercoiling and catenation. It also shows how these enzymes serve as important targets for antibiotics and anti-cancer drugs.

Those links are not abstract. Mismanaged crossings and twists during replication or transcription trigger breaks, stalled forks, altered gene expression, and increased disease risk.

A method that cleanly reads which strand sits on top at each crossing can show when and where the cell’s topological control slips, and it can reveal how drug candidates change those patterns.

Testing the AI tool’s accuracy

To show the software works beyond simple plasmids, the authors tested it on replication intermediates formed in Xenopus egg extracts. This well-established system mimics DNA synthesis outside a living cell.

Protocols for these extracts show they contain all the factors needed to license and copy DNA, which lets scientists pause forks and capture real intermediates.

They also created specific DNA knots and links using E. coli proteins and then tested whether the system could correctly identify them.

The team demonstrated that the method could measure the size of simple DNA circles with about one percent accuracy. It also distinguished between two similar five-crossing knot types by tracking how the strands crossed over or under each other.

“We have done that by developing advanced new image-analysis tools that can do in a matter of seconds that before may have taken hours,” said Pyne.

Telling strands apart

The hardest part of visualizing a tangle is not seeing that two segments cross – it is telling which one is on top. The authors solved this problem by using the small but measurable height difference at each crossing.

The researchers then trained a model to follow the DNA paths without losing track at tight junctions.

“DNA is a really long molecule,” said study co-author Dr. Sean Colloms from the School of Molecular Bioscience at the University of Glasgow. “Just like any long piece of string, the DNA in our cells gets tangled and knotted.”

“At each DNA crossing, we can see which piece of DNA goes over which and this even allows us to tell the difference between one knot and its mirror image, which is important in our studies.”

DNA knots guide new drugs

The ability to classify tangles on single molecules gives drug hunters a readout for compounds that alter replication, transcription, or decatenation.

It can also guide design in DNA nanotechnology, where loops and crossings are part of the scaffold rather than a nuisance.

Previous studies have already used atomic force microscopy to see how supercoiling changes groove width and local recognition. This work suggests that shape cues may help guide protein binding.

“By developing advanced models, we can generate thousands of molecular structures to train future AI frameworks, bringing us closer to visualizing and understanding topology of complex DNA assemblies,” said Dušan Račko of the Polymer Institute of the Slovak Academy of Sciences (SAS).

DNA mapping is not easy

Every method has trade-offs. Atomic force microscopy requires attaching DNA to a surface and pushing gently with a tip, so adsorption and imaging forces can change conformation if conditions are not tuned well.

A 2021 AFM review shows how cations, polymers, and scan settings shape mica results, stressing consistency.

Even so, the Sheffield pipeline reduces observer bias and scales analysis so that tricky cases can be flagged by their confidence scores instead of quietly misread.

It also makes replication biology more concrete by turning fork junctions, gaps, and reversed forks into numbers rather than sketches. That shift from impression to measurement is what lets the field ask sharper questions.

There is room to grow. The same tracing approach should extend to RNA, to protein-nucleic acid complexes, and to engineered lattices.

Adding live imaging and selective chemistries could link topological snapshots to time and to specific proteins, which could refine readouts for topoisomerase drugs and beyond.

The study is published in the journal Nature Communications.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. 

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–