Dyno Phi

Learn/Predict a structure

  1. Sequence
  2. Model
  3. Confidence
  4. Alignment
  5. Complexes
  6. Sampling
  7. Compare

Read the confidence

Every prediction carries its own estimate of where it is wrong. Three readings cover most of it: pLDDT, a per-residue confidence drawn on the structure and along the sequence; PAE, a per-pair error that says whether two parts are placed right relative to each other; and one or two numbers that summarise them for the whole chain or the whole complex.

The ectodomain by ESMFold, coloured by confidence

Loading structure
  • pLDDT ≥ 90 · very high confidence
  • 70 to 89 · confident
  • 50 to 69 · low confidence
  • below 50 · very low confidence

No sequence loaded.

90 and above
Very high. Backbone and side chains can be trusted.
70 to 90
Confident. The backbone is right; side chains may not be.
50 to 70
Low. Treat with care.
Below 50
Very low. Usually disorder, a region with no fixed structure, rather than a wrong fold.

Where the parts sit

Each dark square on the diagonal is a domain the model is sure of on its own. The blocks between them say whether the domains are placed right against each other: dark is trusted, pale is not. Three confident domains with pale blocks between them means each fold is right and the arrangement is a guess.

Hover the grid to read the error between any two residues.

The numbers, and their thresholds

MetricESMFoldAlphaFold 2
pLDDT73.5692.67
pTM0.89440.92
iPAE7.235 Å5.362 Å
pLDDT
Mean per-residue confidence of the predicted complex.
pTM
Predicted TM-score of the whole complex.
ipTM
Predicted TM-score of the interface alone.
iPAE
Mean predicted aligned error across the interface.

Two predictions of the same chain: ESMFold from the sequence alone, AlphaFold 2 with its alignment. For a complex, ipTM above 0.8 means the chains are placed reliably and below 0.5 means they are not; between the two, sample more before deciding.

  • pLDDT ≥ 90 · very high confidence
  • 70 to 89 · confident
  • 50 to 69 · low confidence
  • below 50 · very low confidence

Read more: AlphaFold DB on pLDDT and PAE · ipTM thresholds

Back: ModelNext: AlignmentRead the docs →