PAE can be visualised as a 2D graph, with the protein residues running along the axes. In each square, the shade of green or (blue: low to red: high) in the case of Colabfold, indicates the expected d
Last updated
Was this helpful?
Predicted aligned error (PAE) is a measure of how confident a model is in the relative position of two residues within the predicted structure. PAE is defined as the expected positional error at residue X, measured in Ångströms (Å), if the predicted and actual structures were aligned on residue Y.
On Fastfold for every prediction you will get a default PAE plot, but you can also ask Aura for a custom plot after you have prediction.
Ask Aura:
Plot the PAE of this protein.
Default plots may vary per model type.
PAE is effectively a measure of how confident a model is that the domains are well packed and that the relative placement of the domains in the predicted structure is correct.
A low PAE score between two residues from two different domains means low predicted error. This in turn means the model is confident in the position of these residues. Conversely, a high PAE score means that the model is not confident in their relative position.
Ignoring the PAE score can lead to misinterpretation of the relative position of domains (Guo et al., 2022). One example is the mediator of DNA damage checkpoint protein 1 (AlphaFold ID: AF-Q14676-F1). Its two domains appear to be close together in space, but the PAE indicates that the positions of these domains are essentially random.
The PAE plot will always have a dark green diagonal line running from top-left to bottom-right. This represents residues being aligned against themselves, where confidence is always high by definition, so it is not informative and can be ignored. The biologically-relevant information, in terms of relative orientations, is contained in the regions away from the diagonal.
