A major update to Single Particle Tracking microscopy has been developed by a research team at the University of Cambridge, with big potential for truly understanding the lifecycle of proteins that cause Alzheimer’s and other dementias.

Imagine a screen covered in dots. All the dots are moving. They all look identical. Could you follow one single dot with complete accuracy? Probably not, and neither could the algorithms behind Single Particle Tracking, limiting the applications of this type of microscopy.

A new method, called FidlTrack, developed by Pierre Parrutto at the Avezov Lab has improved the accuracy of single particle tracking. This has opened the door for myriad potential discoveries, with a new window to see inside the living world of cells.

Their paper has been published in Nature Communications: FidlTrack: high-fidelity structure-aware single particle tracking resolves intracellular molecular motion in organelles sensing APP processing

Single Particle Tracking is a powerful technology used to track proteins inside living cells. The scales being viewed are tiny – a cell you see though microscope can contain millions of proteins, of different sizes and shapes, in a state of constant flux.

Exploring the movements of individual proteins in this subcellular world is very complex. For twenty years, there has been a margin of error in single particle tracking, but the UK Dementia Research Institute team from the Avezov Lab at the Department of Clinical Neurosciences, Cambridge University, have worked on algorithms that, for the first time, can accurately track individual proteins in real time.

Showing Alzheimer’s proteins being cleaved

The efficacy of the new method was demonstrated in a variety of cases. An example showed—very clearly, as it happened—how the Alzheimer’s related protein APP (amyloid protein precursor) was cut by the enzyme BACE1 inside the endoplasmic reticulum of a cell.

The endoplasmic reticulum is a folding membrane near the centre of the cell, which has an important function in moving proteins. The imaging Parutto and his colleagues developed is sharp enough to see protein activity right down into sub structures within that compartment, and it can even catch rare, quick movements. This showed precisely when APP was being cleaved.

The new method has a variety of applications: it can show when a nanobody or intrabody binds to a molecule by watching how the molecule’s motion changes. It will also be able to track individual molecules even in very crowded locations such as nerve cell branches. 

FidlTrack is essentially a more precise, more faithful, microscope that will help scientists observe the earliest, hardest to observe steps in protein biology, and provide a live readout of a protein’s state, enabling future discoveries for dementia causing amyloid related proteins and others.

We spoke to Pierre Parutto to ask him about FidlTrack

pierre-parutto-lab-2023_1.jpg

Why did you want to improve SPT?

SPT is very precise because we see individual proteins with good time resolution. But we also lose all the context. You only see dots, and we don't know what else is happening to them. Over the last 20 years SPT has improved a lot: the physical microscopes have gotten better, but the processing and computational part of it has not improved that much. 

Essentially, all the algorithms that do the tracking are mostly the same and have been for 20 years. 

How do you track a protein?

You take a video of a live cell and you have dots that move. These are your proteins. You need to go from this video to an ensemble of trajectories: you want to track one trajectory per molecule.
As an example - if you were to do this with a surveillance video, in footage of humans, you could use the size, body shape, or colour of clothes to help track individuals. But molecules are a challenge because they all look the same. There's very little information available except their position. 

For a long time, researchers could not know the margin for error in their data. You would do an experiment and there would be no way to assess how good the trajectories were - what we call the fidelity. This has been problematic in the field, as SPT could only be used in applications where some errors didn’t matter. 

To open up SPT for new applications, like APP processing, more fidelity is needed. We need to make sure that the trajectory of our protein is not mixed up with the other dots on screen to accurately follow its status.

What specifics helped you solve the problem?

There’s three branches of improvements we explored: why some proteins are harder to track, how to account for the quality of a recording (they can vary), and how to improve tracking.
So, in branch 1, which proteins are harder to track than others? 

Soluble proteins are faster because they just float around. But proteins that are in the membrane of cells or in cellular compartments are slower, move less, and are therefore easier to track. 
We mapped out the landscape of complexity for different proteins by simulating trajectories and evaluated how much tracking error there was.

In branch 2, we knew there would to be a lot of variation in the quality of different recordings. You do an experiment one day - it's very good quality. Another day it might be a bit different because of variation. We needed to estimate the margin for error. To do this, we used ambiguity scoring. It's quite simple: every time the tracking algorithm has to make a choice, every time there are multiple successors (possible for linking), then we're going to consider that there might be an error. This now allows us to have an idea of how often this situation happens in our data, giving us a quantitative measure of data quality.

In branch 3, we explored how to improve tracking fidelity by incorporating a new source of information to the algorithms. 

A lot of the proteins we are most interested in do not live in free space, but are contained in subcellular compartments, like mitochondria, endoplasmic reticulum, the nucleus, and membranes. And these compartment have specific geometries, for example endoplasmic reticulum forms a network and mitochondria are the shape of sausages. Those shapes constrain protein motion: a protein cannot jump between two different mitochondria and has to move along the structure. We demonstrated that adding these constraints to the tracking algorithm improves fidelity.

Do you think this will be a useful new tool for other teams?

We think it's going to be adopted in the field as these methods are applicable to most tracked proteins and are easy to use. One can either use all three elements together or any of those as required. The ambiguity scoring is especially useful and needed to ensure that biological conclusions obtained from Single-Particle Tracking are drawn from high-quality data.

You saw how APP was cut inside the endoplasmic reticulum. How might this help and support Alzheimer's research?

An important pathway in Alzheimer’s is the production of toxic amyloid species resulting from the cleavage of APP. There has been very limited methods for characterising the cleavage status of APP in individual live cells. We now provide such a tool, allowing researchers to characterise the status of APP in live cells. This can be used to characterise how different genetic backgrounds or treatments affect APP cleavage rates. This ultimately could help understanding the biology behind variations in genetic background as well as drug discovery.

Additionally, the technology can be used to assess the status of a wide range of proteins, along Alzheimer’s and other diseases pathways in live cells.

Read the paper:

Parutto, P., Yuan, Y., Davì, V. et al. FidlTrack: high-fidelity structure-aware single particle tracking resolves intracellular molecular motion in organelles sensing APP processing. Nat Commun 17, 2639 (2026). https://doi.org/10.1038/s41467-026-69067-y