Proteins in Motion
A molecule you can see. A molecule you can hear.
“All things are made of atoms, and … everything that living things do can be understood in terms of the jigglings and wigglings of atoms” Feynman et al., 2010, p. 10
Six ways into a moving molecule
Thousands of atoms
One structure
Constant motion
Hollingsworth & Dror 2018, p. 1129“Imagine that an alien lands on Earth, hears about something called a ‘bicycle,’ and wants to understand how it works … Figuring this out given just a picture of a bicycle would be challenging. Watching a movie of someone riding a bicycle would help … A molecular biologist trying to understand how a protein or other biomolecule works faces a similar challenge.”
One · Watch a protein move
A protein is not a picture
Lysozyme, an enzyme in tears and saliva. Nineteen hundred and sixty atoms, held in one shape, surrounded by water.
Shown with the 910 water molecules nearest to the protein.
Two · Can your eyes keep up?
Four things are changing at once
Every number on the right is measured from this simulation. They are the four descriptors the sonification pipeline reads. Each one adds its own mark to the molecule. Watch the molecule and track them all.
Tap a measurement to see only its mark. Tap it again to go back.
Your challenge
Start with the structure alone
Nothing to track yet. Just watch it move.
Three · Hear the protein move
Now listen instead
The same four descriptors, turned into music by the project's pipeline. Instrument, volume, register, and duration each carry one of them.
The same simulation, sonified another way: the three main motions and the cluster, note by note, in the simulation’s own time.
The simulation as music
One hundred nanoseconds of lysozyme become three minutes of music. The molecule keeps step with the score and pauses while a percussion gesture marks each change of cluster.
Alternative sonification
D natural minor60 BPM1 s per ns
How it differs from the main sonification
Main
Alternative
Time
100 ns grouped into nine episodes; it pauses at each change of cluster
The simulation’s own clock: a note every 0.1 ns, nothing grouped or paused
Melody
The same maqām statement in every episode
A new note at every step, its pitch set by PC1
To the ear
Easier to follow
Closer to audification: more faithful to every fluctuation, but harder to follow
PC1 → pitch · PC2 → articulation · PC3 → loudness · cluster → instrument
0:00 / 0:00Live trajectory, in step with the music · drag to turn
Four · How the music is built
Four rules, one maqām
Every cluster plays the same maqām statement. Four things change with the data: the instrument (which cluster), the volume (the dominant motion), the register (side-chain hydrogen bonds), and the duration (how long each episode lasts).
Cluster instrument
Dominant motion volume
Side-chain H-bonds register
Episode length duration
Change the musical system
The same trajectory and the same four rules, rendered in different tuning systems and maqāmāt at 140 BPM.
0:00 / 0:00
Five · Why use sound?
Sonification: what it is and why it helps
“Translation of data into non-verbal sound to facilitate interpretation and communication” UNOOSA, 2023, p. IX
~1600s
Galileo's inclined plane: a ball strikes catgut strings in rhythm.
1980s
Sonification grows as a field.
1992
First International Conference on Auditory Display.
Several sonification techniques took shape from the 1990s onward. These five are the best established.
Audification
The data played directly as a sound wave, with minimal processing.
Auditory icons
Short everyday sounds attached to single events by metaphor, grounded in how we hear our surroundings.
the click of a digital camera.
Earcons
Just as short, but synthetic sounds with no prior meaning, which has to be learned.
a melody signalling battery level.
Parameter mapping PMSonUsed in this project
Data dimensions mapped to sound dimensions: pitch, loudness, duration, timbre. Simple to design, and it conveys information continuously: the most widely used technique.
Model-based
The data become a virtual object that sounds only when you interact with it.
telling a cracked table-tennis ball from an intact one by its sound.
Why parameter mapping needs care
- It is the predominant approach today.
- One ill-judged mapping can make the whole display unusable.
- To count as sonification, a mapping should be neither wholly arbitrary nor too complex, so the relations in the data stay decipherable.
- So much rests on each choice that it needs a systematic design, not an ad hoc one.
♫
Perception
Multidimensional; high temporal resolution; picks signals from noise; works in the background.
◎
Accessibility
Opens dynamic data to blind and vision-impaired researchers.
☼
Engagement
Lets non-experts experience scientific exploration.
∿
Complex data
Large, noisy, temporally variable datasets.
Challenge
Response
Western science is a culture of seeing.
Humans evolved neural circuitry to process large volumes of audio rapidly.
Auditory patterns are recognized less readily.
Visualization also needs training; we just learn it early.
Sound choices can seem subjective or arbitrary.
Accepted standards; a published key or legend for every mapping.
Seen as a “gimmick.”
Involve domain specialists; keep the data's meaning in view.
Intent
Data-driven sound is “necessary but not sufficient”; it must aim to communicate something about the domain.
Inference-preserving
What a listener hears should also hold of the phenomenon.
Hypothesis-generating
Claims concern what a listener can decode, not what the molecule does.
Artistic
Expression; sound may be manipulated freely beyond the data.
Outreach & education
Engage non-experts; spark curiosity; teach.
Scientific research
Sound as analytical medium; mapping grounded in technical accuracy.
← looser coupling to the datacategories overlaptighter coupling to the data →
Earlier molecular-dynamics sonification has sounded single events, tracked bonds chosen in advance, compared states that were already known, or turned molecules into instruments. None follows a whole protein’s state and collective motion together, through one exploratory trajectory.
Earlier MD sonification
Rau et al., 2015
Events in MD visualization: hydrogen bonds forming and breaking are given sounds, to draw attention to events hidden in the picture.
Mitchell et al., 2020
Molecules as musical instruments: live, interactive simulations of 17-alanine, graphene, and a carbon nanotube, turned into sound by scanned synthesis.
Scaletti et al., 2022
Teaching protein folding: simplified lattice-model folding animations, with sound, for the classroom.
Scaletti, Hebel & Gruebele, 2024
Hydrogen bonds during folding: each bond of the WW domain gets its own pitch; listening revealed fast and slow routes through the folding transition.
The two most recent, compared with ours
Babbitt et al., 2026Musicality in protein interaction dynamics informs the multi-scale evolution of prosocial behaviour
Yasar, 2026Sonifying known GPCR functional states
Our approachLysozyme (PDB 1AKI), 100 ns
Starting point
Bound vs unbound interaction
Known GPCR functional states
One exploratory MD trajectory
Molecular variables
Residue vibrations and fluctuation differences
Preselected biological distances and motifs
Conformational clustering, PCA, and hydrogen bonds
Representation level
Residue and site interaction dynamics
Specific functional coordinates
Whole-protein collective dynamics
Main question
Do protein interactions show a music-like acoustic organisation?
Does audio preserve known MD-state information?
Can complex protein dynamics be made interpretable and usable through sound?
Tested with listeners?
No
No
Part of our intended direction
What they focus on
One thing picked or known in advance: a single event, chosen bonds or distances, known states, or a molecule played as an instrument.
What is missing
How to sonify a protein whose important motions are not known in advance. And mappings are rarely tested with listeners other than the researchers who designed them.
How our work differs
We follow one exploratory trajectory of the whole protein, with no known answer to confirm. The cluster it is in and its continuous motion become separate layers of the music. Testing it with listeners is the next step.
Six · What are we asking?
Five research questions
Bringing together computational biology, data analysis, and musical expertise across the university.
For researchers
Behind all this
1
Can you hear a protein move?
How can we turn complex protein motion into sound when we do not already know which movements matter most?
2
How do we turn data into pleasant sound without losing the science?
Can a sonification stay faithful to the molecular data while still being understandable and listenable?
3
What happens when sonification meets Arab and Western musical traditions?
How do musical and cultural choices shape the way scientific data are represented through sound?
4
What does it mean to do interdisciplinary research?
What happens when molecular science, computing, music, culture, and education have to work together on the same problem?
5
What can be gained, lost, or discovered when different forms of knowledge are brought together?
What compromises have to be made, what tensions exist between different fields, how can bringing together different perspectives reshape understanding, and how can these differences be leveraged to generate new insights?