Make PAE plots work on trimmed AlphaFold structures

[phil.cruz@…]

Currently predicted aligned error plots can only be shown with the full length AlphaFold model so that the model has the number of residues matching the PAE matrix size. But AlphaFold models are often trimmed by deleting N and C-terminal residues to match an experimental structure sequence. I'd like to be able to use PAE plots with AlphaFold models that have had residues deleted.

A good way to handle this might be to remove the rows and columns of the PAE matrix corresponding to the deleted residues. ChimeraX knows the full sequence of an AlphaFold model (from PDB SEQRES records) even when residues are deleted, so it can figure out which residues have been deleted and remove those rows and columns from the matrix. Since the residues are deleted, showing predicted errors to those residues is not too useful so removing them from the PAE plot seems sensible.

[Tom Goddard]

Done.

If the AlphaFold model has multiple chains and whole chains are deleted, ChimeraX does not remember those deleted chains and so it will give an error saying the model does not appear to have the same number of residues and the PAE data. So this improvement only allows PAE plots to work with deleted residues as long as whole chains are not deleted. I think the case of deleting whole chains is relatively uncommon.

[Tom Goddard]

Eric tells me that ChimeraX does not remember deleted whole chains. But he suggests that I could look for seqres records in the metadata if I wanted to know about them and handle the case of deleted whole chains.