VanishPoint
Most AlphaFold viewers have a confidence lens: slide a threshold and the model fades away everywhere it is unsure, leaving only the parts it is confident about. It is tempting to read whatever fades first as “wrong” or “not important”. VanishPoint turns that lens into a game — predict where each region vanishes, then watch the lens sweep — and shows that a region can fade early and still be the working heart of the protein. Everything runs locally; nothing about you is collected.
This is the tool. For the regions that fade, your tag is checked against the region’s live UniProt annotations, fetched in your browser. Every region is 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 region guide below still works.
Before you call it: guessing alone would be right about 1 time in 4.
Most first predictions in science miss. Checking is the point.
How to play. Press Start. A region is highlighted in 3D. Predict how confident the model is there — 1 Very high, 2 Confident, 3 Low, 4 Very low. The lens sweeps up and the region fades at its true confidence, scoring your prediction. If it faded early, tag it 1 Broken (a modelling error), 2 Working (disordered, doing a job) or 3 Cut off (removed from the finished protein) — and the live UniProt annotation shows the answer. Press Enter or Next to continue.
Why the faded regions matter
BRCA1's long disordered middle carries binding sites for partner proteins in DNA repair, and c-Myc's disordered front end is its transactivation region, which helps switch genes on. Faded in the model does not mean unimportant: inherited BRCA1 variants raise the risk of breast and ovarian cancer, and MYC is overactive in many cancers.
What the lens does: pLDDT confidence
AlphaFold scores every residue from 0 to 100 for how sure it is of the local shape — a score called pLDDT. The confidence lens simply hides everything below a threshold you choose. EMBL-EBI groups the score into four bands:
- Very high (pLDDT ≥ 90) — a backbone you can usually trust — stays crisp under the lens
- Confident (70–90) — a good guess at the fold
- Low (50–70) — treat the exact shape with caution
- Very low (< 50) — no single fixed shape on its own — fades first under the lens
The one idea to take away. When the lens hides a region, it means AlphaFold is not sure of the region’s exact shape on its own — most often because it is intrinsically disordered, or a flexible linker, or a part that only folds when it binds a partner, or a piece the cell cuts away from the finished protein. 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. None of those means the region is a mistake or unimportant — “flexible and functional”, not “wrong and useless”, is the expert’s starting point. Some of the most important regions in biology — like most of BRCA1, or the front end of c-Myc — fade first.
The regions
Answer key — open it after you play
| Protein | Region | Residues | Confidence | What it does |
|---|---|---|---|---|
| BRCA1 P38398 | N-terminal RING domain | 24–64 | Confident | A small zinc-binding fold that pairs with BARD1 to form the E3 ubiquitin-ligase. The model reads it as confident (mean pLDDT about 86): well determined, though 27 of its 41 residues dip below 90, so it outlasts the faded middle under the lens. |
| BRCA1 P38398 | Central intrinsically-disordered region | 500–1000 | Very low · fades early, still working | Most of BRCA1 is disordered — yet this middle is a working scaffold, carrying binding sites and DNA-damage phosphorylation marks. It vanishes early under the lens, and it is still essential. |
| BRCA1 P38398 | Tandem C-terminal BRCT domains | 1646–1859 | Very high | A phospho-peptide reader: the BRCT pair docks onto partners only after they are phosphorylated. Well folded (mean pLDDT about 91), so it stays crisp under the lens. (A close call: the mean is 91.4, within 2 of the cut at 90.) |
| c-Myc P01106 | N-terminal transactivation region (MYC boxes I–II) | 1–143 | Low · fades early, still working | Disordered, and the business end: it recruits co-activators through the conserved MYC boxes. It reads low (mean pLDDT 50.4; 76 of its 143 residues are below 50), so it fades early under the lens, but this is where the transcription control happens. (A close call: the mean is 50.4, within 2 of the cut at 50.) |
| c-Myc P01106 | C-terminal bHLH-leucine-zipper | 355–439 | Very high | The DNA-binding / dimerisation module. Its helices are locally well defined (mean pLDDT about 93), so it stays crisp under the lens — but it only grips DNA as a MYC–MAX pair. A crisp region can still need a partner to do its job: confidence is about shape, not about working alone. |
| Lysozyme C P61626 | Folded catalytic body | 19–148 | Very high | A compact, disulphide-stapled enzyme that cleaves bacterial cell-wall sugars. Rigid and well determined — it stays crisp all the way down the lens. |
| Lysozyme C P61626 | N-terminal signal peptide | 1–18 | Low · fades early, cut off | The signal peptide that gets the enzyme exported, then is cleaved off. It reads low because it has no fixed fold in the mature protein — a genuinely low-information stretch, correctly faded. |
| Insulin P01308 | C-peptide (a propeptide) | 57–87 | Very low · fades early, 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
The 3D models are fetched live from the AlphaFold Database and coloured by pLDDT; for the regions that fade, the working or cut-off answer is checked against the region’s live UniProt annotations. Each region’s answer band is the band of its mean per-residue pLDDT in the current AlphaFold model, computed at every build from a vendored copy of the model’s own scores (the teaching point is where a region sits on the confidence scale, not an exact number). The misconception it targets — that a faded, low-confidence region is a failure or does not matter — is a documented novice reading of the AlphaFold confidence colouring. Every region is written by hand, not generated, and science-checked. Nothing is stored; no accounts; no servers of ours (your browser talks straight to AFDB and UniProt).
Want to drive the lens on any structure yourself? Open the free AlphaFold viewer. Everything we make is free and non-profit; learn more on the home page.