Fold Commons

HearTheChain

A little game you can play with your eyes shut. Every AlphaFold model comes with a confidence score for each residue (called pLDDT). HearTheChain plays that confidence along the chain as pitch — high notes where the model is sure, low notes where it is not. Listen to a trace, name the protein, then answer the real question: is its low stretch broken, working, or cut off? Nothing is installed; nothing about you is collected.

This is an accessibility-first tool. You can play it entirely by ear and keyboard, and every trace can also be described in words (press D) or shown as a coloured bar chart on reveal — so it works whether you rely on sound, sight, or neither. The full profile of every protein is in the guide below, which works with or without JavaScript.

Loading the game… If it does not start, the full confidence profiles below still work.

Nothing plays until you press a button. Stop (or Esc) silences every sound at once.

How to play. Press Start. A confidence trace plays as a run of notes from the start of the chain to the end. Press R to replay or D to hear it described in words, then press 1–4 to name the protein. One note is about 25 residues and a soft tick marks every 100, so a longer chain plays longer. A game plays 5 of the 8 proteins. If the protein has a low stretch, tag it 1 Broken (a modelling error), 2 Working (disordered, doing a job) or 3 Cut off (removed from the finished protein). There is no way to lose: every round ends with the honest answer.

Why the low trace matters

Tau, the chain whose sound stays low, has no single fixed fold on its own. It binds and steadies microtubules through its repeat region, and in Alzheimer's disease it clumps into tangles inside neurons. A low pLDDT for a chain like this reflects flexibility, not a broken prediction.

What the notes mean: pLDDT confidence

AlphaFold does not just guess a shape — for every residue it also reports how sure it is, a score from 0 to 100 called pLDDT. EMBL-EBI colours the model by four bands. HearTheChain maps the same score to pitch: a very-high residue is a high note, a very-low one is a low note.

The one idea to take away. A low, quiet stretch is not a failed prediction and not an unimportant protein. It means AlphaFold sees no single fixed shape for that region on its own — most often genuine intrinsic disorder, but sometimes a flexible linker between domains, or a part that only folds once it binds a partner. In every case that is an honest report on a dynamic region, which is often exactly where the biology happens. A low stretch can also be a piece the cell cuts off the finished protein, like a signal peptide. Tau is almost entirely low-confidence, and it is one of the most-studied proteins in all of neuroscience. (The reverse also happens: AlphaFold draws part of alpha-synuclein as a confident helix, the shape it takes on a membrane, though on its own it is disordered. High confidence is not proof of a fixed shape.)

The protein profiles

Every trace in the game is the confidence profile of a real protein in the AlphaFold Database, read from the start of the chain to the end: one note per about 25 residues, each the mean confidence of its stretch. The profiles below count the notes in each band — the teaching point is the shape and the length of confidence along the chain, not the exact numbers.

Answer key — open it after you play
ProteinWhat it isConfidence trace (start → end)Low stretchWhy
Tau P10636 A brain protein that binds microtubules, the cell’s internal tracks, and helps keep them steady. Clumps of it are a hallmark of Alzheimer’s disease. 758 residues, 30 notes: 1 low, then 21 very low, then 6 low, then 1 very low, then 1 confident 1–732: working Tau is intrinsically disordered: most of its long chain has no single fixed shape on its own, and it binds microtubules through its repeat region. The low trace is honest about real disorder, not a broken prediction, and certainly not an unimportant protein.
Lysozyme C P61626 A compact enzyme in tears and saliva that cracks bacterial cell walls. 148 residues, 6 notes: 1 confident, then 5 very high 1–18: cut off The signal peptide that gets lysozyme exported, then is cut off. It has no fixed place in the finished enzyme. A small, well-folded enzyme: high confidence almost end to end, lower only over the signal peptide at the start, because its shape really is rigid and well-determined.
Haemoglobin, alpha chain P69905 The oxygen-carrying globin that colours your blood red. 142 residues, 6 notes: 6 very high — A classic globin fold — compact and well determined, so the confidence stays high the whole way along.
Ubiquitin P0CG47 A tiny tag the cell sticks onto proteins to mark them for recycling. (The model here is human polyubiquitin-B: three 76-residue ubiquitin units in a row.) 229 residues, 9 notes: 8 very high, then 1 confident — One of the most stable little folds in biology — confidence is very high all the way across.
p53 (tumour suppressor) P04637 The “guardian of the genome” — its activity is a checkpoint that can halt a damaged cell’s division, for repair or cell death. 393 residues, 16 notes: 1 low, then 2 very low, then 1 low, then 3 very high, then 1 confident, then 4 very high, then 1 very low, then 1 confident, then 1 low, then 1 very low 1–61: working Disordered, yet it is the business end: it binds MDM2 and the transcription machinery. Its floppiness lets it wrap onto several different partners. A mixed trace: p53 has a well-folded DNA-binding core (high) and a small tetramerisation unit (high) flanked by genuinely disordered stretches (low). Low there means flexible-and-regulatory, not wrong.
BRCA1 P38398 A large DNA-repair protein; faults in it raise breast- and ovarian-cancer risk. 1863 residues, 75 notes: 1 confident, then 1 very high, then 2 confident, then 52 very low, then 1 confident, then 9 very low, then 1 confident, then 5 very high, then 3 confident 500–1000: working Two small folded modules near the ends (the RING near the start, the BRCT pair at the very end) joined by a huge disordered middle (very low). The long low stretch is real disorder — most of BRCA1 has no fixed shape on its own.
Serum albumin P02768 About half of the protein in your blood plasma is albumin. It carries fats and medicines around your body. 609 residues, 24 notes: 1 very low, then 22 very high, then 1 confident 1–24: cut off Signal peptide plus a short propeptide: the cell uses them to send albumin out of the cell, then cuts both away. The finished protein in your blood starts after them.
Insulin P01308 A signal your body sends out after you eat. It makes your cells take in sugar. (The model here is the full 110-residue precursor chain.) 110 residues, 4 notes: 1 low, then 2 very low, then 1 low 57–87: cut off The C-peptide links insulin’s two chains while proinsulin folds, then it is cut out. Finished insulin is the A and B chains held together by disulfide bonds. The C-peptide is released alongside it and is studied for roles of its own, but it is not part of insulin.

Methods & about

This is a teaching game, and it is honest about being a simplification. Each note is the mean AlphaFold pLDDT of an equal stretch of about 25 residues, computed at every build from a vendored copy of the model's own per-residue scores, not the exact per-residue values; the confidence bands follow EMBL-EBI's own colour cut-offs (≥90 / 70–90 / 50–70 / <50). The misconception it targets — that low confidence means the model is wrong or the region does not matter — is a documented novice reading of the AlphaFold confidence colouring; the honest counter is that low pLDDT reports a region with no single fixed shape on its own (most often intrinsic disorder, but also flexible linkers or fold-on-binding parts), which is frequently functional, or a piece the cell cuts away. Rarely is not never: the model can also be unsure about a part that is actually ordered, for example when it has few related sequences to learn from. Every description is written by hand, not generated; every number is read from the model. The game keeps no score history, uses no accounts, and sends nothing anywhere.

Want to see and rotate these models in 3D? Open the free AlphaFold viewer. Everything we make is free and non-profit; learn more on the home page.