A Chip That Watches Single Proteins Change Shape Could Change How We Design Drugs
Your body contains roughly 20,000 different kinds of proteins. Each one is a tiny machine that folds, twists, and flexes thousands of times per second to do its job: digesting food, firing neurons, fighting infections. When one of those machines jams or snaps into the wrong shape, you get disease. But for decades, watching a single protein change shape in real time has been like trying to film a hummingbird’s wings with a pinhole camera. The motions are too fast, the proteins are too small, and the measurement tools introduce too much noise.
Researchers at the University of Queensland just solved that problem. In a paper posted to arXiv on July 17, they describe a silicon-chip sensor that can track the shape changes of a single protein molecule at sub-microsecond speeds, continuously, for minutes at a time. No fluorescent dyes, no averaging over millions of molecules, no physical tether that might alter the protein’s behavior. The device is a nanoscale stethoscope pressed up against a single molecule, and it’s revealing things about protein motion that nobody had seen before.
Background & Context
Proteins aren’t static objects. They breathe. Their backbones twist, their side chains rearrange, their domains open and close like tiny jaws. These motions are not just physical curiosity. They control how enzymes catalyze reactions, how receptors transmit signals across cell membranes, and how antibodies recognize their targets. A drug designer who understands exactly how a protein wiggles can design a molecule that wedges into the right crevice at the right moment.
The problem is measurement. The standard approach attaches a fluorescent dye to the protein, a molecular lightbulb that blinks when the protein moves. It works, but the dye itself is often larger than the motion being measured, and it alters the very dynamics you’re trying to see. Worse, fluorescent dyes bleach out after a few milliseconds at high speed, forcing researchers to choose between seeing fast motions and watching for long periods. Label-free methods exist, but they’ve been constrained by a subtler enemy: Brownian motion. A protein floating in water jiggles randomly, and in existing sensors, that jiggling shows up as noise that completely buries the signal from structural changes.
What the Researchers Did
The Queensland team, led by Warwick Bowen, took a different approach to the noise problem. Instead of fighting Brownian motion, they made their sensor blind to it.
They built a slotted photonic crystal cavity: a silicon chip with a 55-nanometer-wide gap etched into it. Light pumped into this gap creates an optical field so uniform that a protein can drift around inside without changing the signal. If you’ve ever watched an autofocus camera hunt for focus as a subject moves, you’ll appreciate the difference: the Queensland team built a sensor with infinite depth of field at the molecular scale. Their simulations show that Brownian motion noise is suppressed by a factor of 60 compared to the plasmonic sensors used in previous single-molecule experiments.
The second clever trick was how they held the protein in place. Instead of chemically gluing it to the surface (which would denature it), they let a thin “hard layer” of proteins form naturally on the cavity walls. Individual proteins then get loosely trapped in this soft molecular blanket, close enough to be monitored but free enough to maintain their natural shape and motion. It’s the difference between pinning a butterfly to a board and watching one rest on your hand.
The sensor is fabricated on silicon wafers using the same lithography techniques that make computer chips, and individual cavities are detached from the wafer using a tapered optical fiber. The fiber both delivers the laser light and reads out the reflected signal, making the device compact enough to dip into a droplet of protein solution.
What They Found
The team tested their sensor on ferritin, an iron-storage protein found in nearly every living organism. Ferritin is built like a hollow soccer ball: 24 protein subunits form a spherical cage roughly 12 nanometers across, with tiny channels that let iron atoms pass in and out. These channels are thought to open and close through conformational changes, making ferritin an ideal test subject.
The sensor didn’t just see motion. It saw distinct, discrete steps between two states. Ferritin flips back and forth between an “upper” and “lower” conformation, with the optical signal showing a clear two-state system.
| Measure | Holoferritin (with iron core) | Apoferritin (empty shell) |
|---|---|---|
| Mean dwell time, upper state | 102.3 ± 0.7 μs | 139 ± 4 μs |
| Mean dwell time, lower state | 134.8 ± 0.9 μs | 174 ± 5 μs |
| Modal transition path time | 5.3 ± 0.5 μs | 2.1 ± 0.3 μs |
| Fastest transition observed | — | ~400 ns |
| Transitions in single trace | 71,718 (8.5 s) | 12,767 (1 s) |
| Potential well type | Double well | Double well |
| Memory effects in dwell times | No (exponential) | Yes (heavy-tailed) |
The numbers are staggering for anyone who works with single-molecule biophysics. In an 8.5-second trace of a single holoferritin molecule, the sensor captured over 71,000 discrete transitions between states. Previous label-free methods couldn’t even resolve individual steps, let alone count them by the tens of thousands.
Apoferritin (the empty-shell version) turned out to be less stable and more interesting. Its dwell times follow a heavy-tailed distribution characteristic of systems with long-term memory, meaning the protein “remembers” its structural history. The state it was in a millisecond ago influences how long it stays in its current state. Holoferritin, with its iron cargo bracing the structure from the inside, behaves more like a memory-less system. The iron core literally stiffens the protein’s personality.
The transition path times (the actual duration of the shape change) are shockingly fast. Some apoferritin transitions clock in at 400 nanoseconds. That’s 0.0000004 seconds. The sensor’s bandwidth extends above 1 MHz, roughly 100 times faster than any previous label-free single-protein measurement.
Why It Matters
Ferritin isn’t just an academic test case. When ferritin malfunctions, the consequences are devastating. Mutations in the ferritin light chain cause neuroferritinopathy, a rare neurodegenerative disorder where iron accumulates in the brain’s basal ganglia, leading to movement disorders and dementia. Other variants cause hypoferritinemia, where patients can’t store enough iron. Understanding exactly how ferritin’s conformational gating works, which this sensor can now reveal, is a direct path toward therapies that fix broken channels.
The applications extend well beyond ferritin. The sensor platform is general-purpose. Any protein that undergoes conformational changes could, in principle, be studied this way. Drug companies spend billions screening molecules against protein targets without knowing which conformation the target is actually in. A sensor that watches a protein’s shape in real time could tell drug designers when the binding pocket is actually open.
Ferritin nanocages are also being developed as drug delivery vehicles: hollow protein shells that can be loaded with chemotherapy drugs and directed to tumors. The release kinetics depend entirely on the cage’s conformational dynamics. Watching the channels open and close at their native speed tells engineers exactly when and how the cargo will leak out.
How It Could Change Our Lives
Imagine a future version of this technology shrunk onto a disposable chip. A blood sample from a patient with a suspected protein-misfolding disease (Alzheimer’s, Parkinson’s, ALS) flows over an array of these sensors. Each sensor grabs a single copy of the suspect protein and watches it move. A protein that spends too much time in a misfolded state flashes a warning. A drug candidate that nudges it back toward the healthy conformation lights up green on the same chip.
That’s not tomorrow. But the Queensland team’s sensor is fabricated on silicon using standard chip-making processes, and they’ve already demonstrated that individual cavities can be detached from wafers containing thousands of identical devices. The path to high-throughput, multiplexed single-molecule screening is visible. The paper explicitly notes that “the sensors are lithographically fabricated in arrays of hundreds of devices and are fibre-integrated, making them naturally suited to scalable high-throughput measurements such as parallelized protein detection and drug screening.”
The practical timeline depends on noise improvements. Right now, protein motion dominates the signal below 1 MHz; the team estimates they could suppress residual Brownian motion by another factor of 100 by positioning proteins at the slot’s center, where the field is 30 times more uniform. Above 1 MHz, silicon’s thermo-optic noise takes over; switching to silicon nitride could buy another factor of 10. Together, these improvements would push temporal resolution into the nanosecond regime, where most protein chemistry actually happens.
The Bigger Picture
Biophysics has spent decades developing ever-more-elaborate ways to look at proteins. X-ray crystallography gives static snapshots. Cryo-EM gives ensemble averages of frozen states. NMR gives dynamics but requires milligrams of purified protein. Fluorescence gives single-molecule trajectories but for milliseconds before the dye dies.
The Queensland sensor doesn’t replace any of these. It adds a fundamentally new capability to the toolkit: continuous, label-free observation of a single protein over physiologically relevant timescales at its actual speed of motion. The paper lists questions this could help answer: whether protein dynamics are ergodic (do they explore all possible shapes over long times?), how hidden states govern transitions, and what role rare pathways and long-lived intermediates play in function. These aren’t incremental questions. They’re the ones that textbooks still get wrong.
Limitations & What’s Next
The sensor currently lacks molecular specificity. A wavelength shift tells you that something changed, but not which atoms moved. The ferritin results are convincing because they match predictions from simulations, because they respond to ionic strength and iron content exactly as expected, and because control proteins (catalase, BSA) don’t show the same stepping behavior. But to identify specific structural changes (which loop moved, which hinge opened) you’d need to combine the sensor with molecular dynamics simulations or complementary structural techniques.
The soft layer approach is elegant but comes with a tradeoff. Proteins loosely bound to a hard corona might not behave exactly like free-floating proteins in cytoplasm. The team acknowledges this and suggests future work should quantify the perturbation.
The most exciting path forward is the one they sketch at the end: arrays of these sensors, each watching a different protein, reading out conformational dynamics in parallel. Silicon photonics has followed the trajectory that silicon electronics blazed decades ago: smaller, cheaper, more parallel. A thousand-channel single-molecule conformational sensor might not be science fiction for very long.
📄 Source: Sung, H., Scholten, S. C., Sasheva, P., Marinković, I., & Bowen, W. P. (2026). Sub-microsecond conformational dynamics in an optical nanocavity. arXiv:2607.15925v1.