Single-Molecule Techniques: How We See & Touch Molecules

Single-molecule techniques let researchers watch and physically manipulate one molecule at a time, revealing behaviors that bulk measurements average away. Where traditional biochemistry works with billions of molecules at once and reports their collective average, these methods can track how a single protein folds, measure the force needed to unzip a single strand of DNA, or follow one enzyme as it pauses mid-reaction. The toolkit spans from optical tricks that make individual fluorescent labels glow bright enough to see, to tiny mechanical probes that grab a molecule and pull on it with precisely calibrated force. Together, these approaches have reshaped how we understand molecular biology, and they continue to open doors into territory that was simply invisible a generation ago.

Seeing One Molecule at a Time

The most widely used way to “see” a single molecule is to attach a fluorescent label to it and detect the photons it emits. A standard fluorescent dye emits a limited number of photons before it permanently bleaches, so the engineering challenge is collecting enough light, fast enough, before the signal disappears. Modern detectors and careful optics have made this routine in well-equipped labs. One particularly powerful variant is single-molecule FRET, often called smFRET, which measures the distance between two fluorescent labels on the same molecule. When the labels are close together, energy transfers efficiently from one to the other; when they drift apart, the transfer drops. This distance sensitivity makes smFRET a molecular ruler for tracking shape changes in proteins and nucleic acids in real time.¹

Conventional optical microscopy hits a resolution wall at roughly 200 nanometers because of the physics of light diffraction. Super-resolution techniques break through that wall by exploiting the fact that you can localize a single glowing point far more precisely than the blur it produces. Stochastic optical reconstruction microscopy, known as STORM, works by switching fluorescent labels on and off at random so that only a sparse handful glow in any given camera frame. Each glowing spot is localized to within a few tens of nanometers, and the composite image built from thousands of frames can resolve structures well below the diffraction limit.² This approach earned part of a Nobel Prize and has since become a workhorse for imaging the fine architecture of cells.

At cryogenic temperatures, the game changes further. Single organic molecules cooled on atomically flat surfaces can emit light with extremely sharp spectral lines, known as zero-phonon lines, because thermal vibrations are largely silenced. Researchers have demonstrated this with individual terrylene molecules sitting on a hexagonal boron-nitride surface, producing spectral fingerprints clean enough to distinguish one molecule from another based purely on its emission frequency.³ While this kind of experiment lives more in physics labs than in biology, it illustrates just how far single-molecule detection can be pushed under ideal conditions.

Touching Molecules with Force

Seeing is only half the story. The other half is physically grabbing a molecule and applying controlled force to it. Atomic force microscopy, or AFM, does this by pressing a cantilever with an ultrasharp tip against a surface-bound molecule. In force spectroscopy mode, the cantilever pulls on a protein or nucleic acid tethered between the tip and a surface, stretching it until bonds break or domains unfold. Early force-clamp experiments on titin, the giant elastic protein in muscle, showed that individual protein modules unfold in discrete steps under constant pulling force, and that the time each module waits before unfolding follows an exponential distribution, a hallmark of a random, thermally activated process.⁴

More recently, improvements in cantilever design have pushed AFM’s sensitivity and time resolution into impressive territory. Focused ion beam-modified cantilevers can now detect molecular states lasting as briefly as one millisecond, with force precision good enough to pick up the subtle effects of pH changes on protein-folding landscapes.⁵ That kind of temporal resolution means AFM can now catch fleeting intermediate states that older instruments would have missed entirely.

Optical tweezers take a different approach to touching molecules. A tightly focused laser beam creates a gradient of light intensity that can trap a small bead, typically made of glass or polystyrene, in three dimensions. When a molecule is tethered between the bead and a surface (or between two beads), the experimenter can stretch, twist, or relax it while measuring forces with sub-piconewton sensitivity. The trap’s stiffness is calibrated by analyzing the Brownian motion of the bead, fitting its power spectrum to extract the trap’s restoring force constant.⁶ Calibration turns out to be quietly critical. Temperature changes and proximity to surfaces alter the viscosity the bead experiences, and if those effects are not corrected, force measurements can be significantly off.⁷

Mapping Energy Landscapes One Pull at a Time

When you pull on a protein hard enough to unfold it, you are not just breaking it. You are probing the energy landscape it lives in, the series of hills and valleys that govern how it folds, misfolds, or stays stable. By repeating pulling experiments hundreds or thousands of times and recording the forces at which unfolding occurs, researchers can reconstruct key features of that landscape: the height of the energy barrier separating the folded and unfolded states, and where along the stretching coordinate the transition state sits.

Work on the small protein barnase illustrates this well. By combining pulling experiments (where force increases until the protein unfolds) and passive experiments (where the molecule is held near its equilibrium and allowed to fluctuate between states), researchers extracted both the unfolding and folding kinetic rates and found good agreement with earlier bulk measurements.⁸ An interesting finding from these experiments is that the position of the transition state is not truly fixed. While it stays at a roughly constant distance from the folded state, it shifts relative to the unfolded state as force increases, meaning the landscape itself warps under tension.⁹ That detail matters because the standard analytical models used to interpret pulling data assume the landscape is static. When the range of applied forces is large enough, those models start to diverge from each other, and researchers need more sophisticated frameworks to get reliable numbers.¹⁰

Watching Molecular Machines Do Their Jobs

Some of the most dramatic applications of single-molecule techniques involve watching molecular machines in action. RNA polymerase, the enzyme that reads DNA and produces RNA transcripts, has been studied extensively with optical tweezers. Researchers can attach beads to the polymerase and the DNA template, then watch the enzyme chug along while measuring how fast it moves and when it stalls.

These experiments have revealed that RNA polymerase does not move at a steady pace. It encounters specific DNA sequences that cause it to pause, and the pausing process unfolds in stages. Almost all polymerase molecules slow down when they hit a pause site, with pausing efficiencies above 70 to 85 percent regardless of how much force is applied to assist or oppose the enzyme. But the duration of pauses is strongly force-dependent, meaning you can tease apart the mechanical steps of pausing by varying the load.¹¹ The enzyme does not immediately start sliding backward (backtracking) when it pauses. Instead, it enters an initial paused state and then, over the course of one to ten seconds, may slip into a deeper, longer-lived backtracked state. This two-step mechanism would be invisible in any bulk experiment, where you would only see an average slowdown.

Single-molecule approaches have similarly transformed our understanding of DNA-processing enzymes. Topoisomerase V, an enzyme that relieves the torsional stress in supercoiled DNA, was shown through micromechanical experiments to relax DNA by releasing multiple turns at once through a constrained swiveling mechanism, powered by the torque stored in the supercoiled DNA and limited by friction between the protein and the DNA strand.¹² ABC transporters, the molecular pumps that shuffle molecules across cell membranes, have also yielded surprises. Cryo-electron microscopy snapshots combined with single-molecule dynamics data showed that ATP binding alone, without the energy-releasing step of hydrolysis, is enough to power the conformational change that moves cargo across the membrane.¹³

When Enzymes Fluctuate

Textbook enzymology treats each copy of an enzyme as identical, cranking through reactions at a fixed rate. Single-molecule experiments have complicated that picture. When you track an individual enzyme’s catalytic activity over time, you often see its rate fluctuate, sometimes catalyzing reactions quickly and sometimes slowly, even under constant conditions. This phenomenon, called dynamic disorder, arises because the enzyme slowly shifts between different conformational states, each with its own catalytic speed. These conformational fluctuations happen on timescales of milliseconds to seconds, much slower than a single catalytic turnover but fast enough to influence the enzyme’s output.¹⁴ The practical consequence is that in a cell, the same enzyme molecule can behave very differently from moment to moment, adding a layer of noise to cellular chemistry that bulk measurements would flatten into a single average rate.

Reading DNA Through a Tiny Hole

Nanopore sequencing represents one of the most commercially successful offshoots of single-molecule science. The concept is elegantly simple: thread a DNA strand through a pore just wide enough for one nucleotide at a time, and measure how the ionic current through the pore changes as each base passes through. Different bases block the current differently, producing a signal that can, in principle, be decoded into the DNA sequence.

The physical mechanism behind the current blockade turns out to be more nuanced than the “different-sized bases block the pore differently” picture might suggest. In biological nanopores like MspA, simulations have shown that the dominant factor is how much water each nucleotide displaces from the pore constriction. The steric and base-stacking properties of the nucleotides determine how much water is pushed out, and the water content in turn sets the ionic current.¹⁵ In solid-state nanopores, computational studies have shown that at least three consecutive nucleotides influence the blockade current at any given moment, and single-base substitutions can shift the current by as much as 25 percent in ultrasmall pores.¹⁶ That three-base context dependency is both a challenge (the signal from each position depends on its neighbors) and an opportunity (it provides redundancy that computational methods can exploit).

Amplifying the Signal with Gold

One persistent limitation of single-molecule fluorescence is that individual dye molecules simply do not emit many photons per unit time. Gold nanostructures can dramatically change this. When a fluorescent molecule sits near a gold nanoparticle or nanorod, the metal’s surface plasmon resonance can amplify the local electromagnetic field, boosting the molecule’s fluorescence by orders of magnitude. Single gold nanorods have been shown to enhance fluorescence by up to about a thousandfold when the dye’s spectrum is tuned to match the rod’s plasmon resonance.¹⁷

Pairs of gold nanorods arranged end to end push the effect even further. In two-photon excitation experiments, gold nanorod dimers have yielded single-molecule fluorescence enhancement factors on the order of a hundred million, making it possible to detect individual dye molecules under conditions that would normally require far higher concentrations.¹⁸ Optimized monocrystalline gold nanoparticles have also enabled real-time single-molecule studies with microsecond temporal resolution, detecting more than ten million photons per second from a single fluorophore and bringing single-molecule correlation spectroscopy down to sub-second acquisition times.¹⁹ These are not just laboratory curiosities. Plasmon-enhanced detection could eventually make single-molecule assays fast enough and bright enough for routine diagnostic use.

Following Molecules Inside Living Cells

All the techniques described so far originated in the controlled environment of a glass slide. Increasingly, single-molecule methods are being brought into living cells, where molecules move through crowded, dynamic environments that no in-vitro system fully captures. Single-particle tracking, or SPT, follows individual labeled proteins as they diffuse, get confined, or are actively transported within a cell.²⁰ Different motion behaviors reflect different biological states: a receptor diffusing freely on the cell membrane behaves differently from one that has bound its ligand and is being internalized.

A persistent challenge with live-cell tracking is that fluorescent labels eventually bleach, limiting observation windows. A recent approach uses weak-affinity protein labels that continuously exchange between labeled and unlabeled states, effectively refreshing the fluorescent signal and extending tracking times to 20 seconds or more per molecule. Using this strategy on the epidermal growth factor receptor (EGFR), researchers measured a drop in diffusion coefficient from about 0.25 to 0.16 square micrometers per second after stimulation with growth factor, along with an increase in the fraction of immobile receptors, consistent with receptor activation and clustering.²¹ Live-cell single-molecule tracking of transcription factors in the nucleus has similarly provided precise measurements of how these proteins search for and bind their target DNA sequences, complementing genome-wide sequencing approaches that rely on chemical cross-linking and can introduce artifacts.²²

Scaling Up with Parallelism

A persistent criticism of single-molecule experiments is that they are slow. Watching one molecule at a time means you need many individual experiments to accumulate statistical power. Several engineering solutions address this. Zero-mode waveguides, tiny holes in a thin metal film with diameters smaller than the wavelength of light, confine the observation volume to zeptoliters, small enough that only one or a few molecules contribute signal at a time even at micromolar concentrations. Arrays of these waveguides enable highly parallel single-molecule observation with microsecond temporal resolution.²³ This is the technology underlying commercial single-molecule real-time (SMRT) DNA sequencing, which produces long reads with minimal sequence-composition bias. Adding nanopores to the waveguide floor has boosted loading efficiency by five orders of magnitude and made loading nearly independent of DNA fragment length, a step toward sequencing from vanishingly small input quantities.²⁴

Acoustic force spectroscopy offers parallelism on the mechanical side. Instead of pulling on one molecule with a single optical trap or AFM tip, acoustic waves apply force simultaneously to hundreds or thousands of microbeads tethered to a surface, each attached to its own molecule.²⁵ Because each bead is tracked independently, you get many single-molecule measurements in a single experiment, dramatically improving throughput for studies of bond strength and molecular elasticity.²⁶

Machine Learning Is Changing How the Data Gets Analyzed

Single-molecule experiments generate enormous volumes of noisy data. A typical smFRET experiment produces thousands of fluorescence traces, each of which needs to be classified as usable or not, then segmented into distinct states, and then fitted to a kinetic model. Historically, much of this was done by hand or with semi-automated tools that required extensive tuning. Deep learning is now taking over large parts of this pipeline. A tool called AutoSiM uses neural networks to automatically select high-quality single-molecule traces, achieving roughly 90 percent agreement with expert manual selection while working far faster.²⁷

A more comprehensive suite called Deep-LASI handles the full analysis workflow for multi-color smFRET experiments: sorting traces, determining correction factors, and classifying dynamic state transitions, all in roughly 20 to 100 milliseconds per trajectory.²⁸ The speed matters not just for convenience but because it enables iterative experimental design. If you can analyze your data in near-real time, you can adjust conditions mid-experiment rather than discovering problems hours later during manual analysis.

The Quiet Importance of Surface Preparation

Ask anyone who runs single-molecule experiments what the hardest part is, and the answer is often not the optics or the analysis. It is preparing the surface. Most single-molecule fluorescence experiments require molecules to be tethered to a glass slide, and the slide’s surface must resist nonspecific sticking of proteins, DNA, and fluorescent contaminants. The standard approach coats slides with polyethylene glycol (PEG), a polymer that creates a brush-like layer repelling unwanted molecules. The traditional protocol involves multiple chemical steps spread over several hours.

Recent work has simplified this considerably. A one-step PEG-silane method grafts the passivation layer directly to glass in minutes rather than hours, with equal or better performance in reducing nonspecific interactions and at lower reagent cost.²⁹ Further refinements have continued to streamline the process, improving reproducibility and making single-molecule experiments more accessible to labs without years of surface-chemistry expertise.³⁰ Surface preparation might seem mundane, but it is the difference between clean data and an afternoon spent troubleshooting mysterious fluorescent spots that turn out to be stuck contaminants.

Sensing Single Spins with Diamond Defects

The techniques above all operate on the scale of biology: proteins, DNA, and their mechanical and optical properties. A very different single-molecule frontier uses quantum physics. The nitrogen-vacancy (NV) center in diamond is a point defect where a nitrogen atom and an adjacent vacancy in the crystal lattice form a system whose electron spin can be initialized, manipulated, and read out optically at room temperature. This makes it a remarkably sensitive magnetometer. By placing an NV center close to a diamond surface, researchers have detected the nuclear magnetic resonance signal from a tiny volume of organic material sitting outside the diamond, sensing the nanotesla-scale magnetic fluctuations produced by the protons in the sample.³¹ This amounts to NMR spectroscopy at the nanometer scale, a feat that conventional NMR machines, which need macroscopic samples, cannot approach. The physics of the NV center and the techniques for working with single spins have been developed extensively over the past two decades.³² Whether this will eventually allow imaging of individual protein structures at atomic resolution remains an open question, but the trajectory of improvement has been steep enough to keep a large community of physicists and engineers working on it.