Structure Prediction¶
Task designed for prediction a protein’s 3D structure from amino acid sequence with ColabFold or Openfold services.
Output¶
Predicted Aligned Error (PAE) reports AlphaFold’s expected position error of predicted residue regarding the structure aligned. PAE is useful for assessing confidence in global features (molecule interface), especially domain packing. Low PAE suggests AlphaFold is confident about the relative domain positions. Meanwhile, high PAE values suggest the relative positions of the domains should not be interpreted.
PLLDT scores plot shows the pLDDT score per position for each of the five AlphaFold models predicted.
AlphaFold produces a per-residue estimate of its confidence score (pLDDT) ranging from 0 to 100.
- pLDDT > 90: residues are expected to be positioned with a high accuracy. These models should be suitable for any application that benefits from high accuracy (e.g. characterising binding sites).
- 70 < pLDDT < 90 residues are expected to be positioned reasonably well (generally, a good backbone prediction).
- 50 < pLDDT < 70 corresponds to residues modelled with low confidence, which should be treated with caution.
- pLDDT < 50 indicates residues that should not be interpreted. Such scores are a reasonably strong predictor of disorder, i.e. they suggest that the region is either unstructured in physiological conditions, or gets structured only as part of a complex.
Sequence coverages plot shows the number of homologous identified across the representative sequence and colored by the sequence identity to query.
References