Wiki · Measurement & Inference · concept
Single-molecule biophysics
Watching and measuring one molecule at a time instead of averaging over billions.
Single-molecule biophysics measures one molecule at a time. A classical experiment reports the average behavior of billions of molecules in a test tube — useful, but it blurs everything together. Single-molecule methods instead watch a lone actor: one motor protein stepping along a filament, one DNA strand being unzipped, one enzyme opening and closing. You see the individual events, not the smeared-out mean.
How it works
The trick is to grab or light up a single molecule and follow it in real time. Optical tweezers use a focused laser to hold a microscopic bead, and magnetic tweezers pull on one with a magnetic field — tether a molecule between beads or a surface and you can stretch it, twist it, and feel it push back. Single-molecule fluorescence, and especially FRET, tags a molecule with dyes and reads out its shape changes as flickers of color. Force spectroscopy pulls on a molecule to map how it unfolds. What you record are forces (piconewtons), displacements (nanometers), and dwell times — and because it’s one molecule at a time, you get whole distributions, not just averages.
Why it matters (for bio × AI)
This is the ground truth that AI models are ultimately judged against. You can predict a structure, but only measurement tells you how the molecule actually moves, how hard it pulls, and how tightly it binds. Single-molecule data is the quiet engine behind believable models: it exposes the dynamics and the molecule-to-molecule heterogeneity that a single static structure can never show. A picture of one frozen pose is not the same as watching the machine run.
Related
See AlphaFold for the prediction side of the same coin — structure guessed from sequence, versus behavior measured one molecule at a time.