BioPython DSSP keyerror for a specific PDB file (1YYJ)

Question

[sajid]

I am extracting secondary structural information from PDB files using DSSP with Biopython. For a specific PDB file (1YYJ), I am getting KeyError: ('A', (' ', 80, ' ')). Minimum code for reproducing the error is as follows:

from Bio.PDB import

parser = PDBParser(QUIET=True)
structure = parser.get_structure(id='struct',file="1YYJ.pdb") 
model = structure[0]
dssp = DSSP(model,"1YYJ.pdb")
dssp["A",(" ",80," ")]

I am getting (" ",80," ") by using get_id() method of Bio.PDB.Residue module. I did some debugging and found that keyerror does not occur upto (" ",79," "). Using the debugging code below, I noticed that, dssp is missing residue 80 (79 and 81 present) for some weird reason, and after 80, dssp entries are not consistent with PDB parser entries. The alignment between the two seems to be broken after the missing 80th residue.

residues = Selection.unfold_entities(model['A'], 'R')
for residue,element in zip(residues,dssp):
    print(residue.get_id(),element)

My script is working fine on around 2000 PDB files, but producing this weird behavior for 1YYJ. Please help me. Thanks a lot.

Answer 1

[Mensur Dlakic]

Two residues (80-81) are missing backbone atoms, which is a chain break. It may show up differently when you parse the sequence because some backbone and sidechain atoms are present for those residues, but for DSSP that's a chain break. It seems that DSSP.py also completely skips over those residues.

A fix for this particular structure is to use the model #2, where the backbone seems to be complete. That's not a long-term solution, though, because you will definitely encounter other structures with chain breaks. My recollection is that SCWRL4 can fix missing backbone atoms, in addition to its main purpose of optimizing sidechains.

Answer 2

[sajid]

Thanks a lot for this pointer. I will try to figure something out using your direction.