HingeHunt
AlphaFold’s per-residue confidence (pLDDT) can be high everywhere and the model can still be unsure how the pieces fit together. That second question — how confident is the model about where each domain sits relative to the others — is answered by a different map: the predicted aligned error (PAE). Judge each protein’s domain arrangement, then reveal the real PAE and find the hinges the confidence colour hides. Everything runs locally; nothing about you is collected.
This is the tool. The 3D model is coloured by pLDDT and looks crisp throughout. Your call — confident or uncertain arrangement — is scored against the live PAE matrix fetched from AlphaFold, not a stored answer. The proteins and their domains are listed in full below (works without JavaScript). A native version, when it ships, adds offline use; it never gates the web.
Loading the game… If it does not start, the full protein guide below still works.
Before you call it: guessing alone would be right about 1 time in 2.
Most first predictions in science miss. Checking is the point.
How to play. Press Start. Look at the crisp 3D model and judge the domain arrangement — Confident or Uncertain (buttons or keys 1/C, 2/U). The real PAE map is revealed and your call is scored against the measured inter-domain error. Press Enter or Next to continue.
Why the gap between the domains matters
Calmodulin's two lobes are joined by a flexible linker. When calcium binds, the lobes can wrap around a target helix from two sides, so there is no single correct distance between them, and a high PAE between the lobes is the honest answer. PGK1's two domains swing closed on their substrates during each reaction cycle. Each domain on its own can still be predicted with very high confidence.
pLDDT vs PAE — two different questions
pLDDT asks: for each residue, how sure is the model about the shape right here? PAE asks: if you line the model up on residue i, how far off is residue j expected to be? Big off-diagonal PAE between two domains means the model is unsure how those domains are arranged relative to each other — a hinge — even when every residue’s local shape (pLDDT) is crisp. That gap is a common misreading, and it is exactly what this game makes you look for.
The proteins
Answer key — open it after you play
| Protein | Domains | Expectation |
|---|---|---|
| Calmodulin CALM1 · P0DP23 | N-lobe (1–75), C-lobe (82–148) | Uncertain — the two lobes swing on a flexible linker. |
| Phosphoglycerate kinase 1 PGK1 · P00558 | N-domain (1–185), C-domain (200–417) | Confident — the hinge opens and closes in life, but the model is sure of the one arrangement it drew (mean inter-domain PAE about 5.5 Å). |
| Myoglobin MB · P02144 | Single globin domain (1–154) | Confident — one compact domain, nothing to arrange. |
Methods & about
This is a teaching game, honest about being a simplification. The score is computed from the live AlphaFold PAE matrix (mean predicted aligned error over the off-diagonal blocks between the listed domains): under 10 Å reads as a confident arrangement, over 15 Å as an uncertain hinge, in between as partly uncertain. Domain boundaries are approximate (drawn from UniProt/family annotations) and the thresholds are teaching cut-offs, not official limits. Structures + PAE are the real AlphaFold predictions fetched in your browser. The game keeps no score history, uses no accounts, and sends nothing anywhere.
Want the deeper experience? See the Fold Commons apps. Everything we make is free and non-profit; learn more on the home page.