mis-association (misalignment) between structure 1ml1 and its own sequence

[Elaine Meng]

If I open 1ml1 and then click the “monotim” to show all the chains associated with the sequence, there are two blocks of mismatches near the beginning of the sequence, but if you mouse over you can see it is instead an issue of the sequence and structure being shifted over by one. If I just click the “A” link, I also get two blocks of mismatches just for that one, but shorter. Screenshot of sequence windows attached.

[Eric Pettersen]

Did the mmCIF entry for 1ml1 change? The chain name is now "triosephosphate isomerase" instead of "monotim", though the sequences look to be the same. The mis-alignments also no longer exist.

[Elaine Meng]

I see, this only happens with the PDB file. I tried the one I had from May 2 and then also fetched a new one (ignore cache) and it still happened. Use: open 1ml1 format pdb

[Elaine Meng]

Oops I didn't ignore cache. When I actually ignore cache I get an error fetching (any) PDB but not mmCIF. Is RCSB no longer providing PDB? Or is this the same problem mentioned in our last dev meeting? Let me know if it needs a new ticket. E.g.

open 1zik format pdb ignoreCache true
Fetching url ​http://www.pdb.org/pdb/files/pdb1zik.ent failed: HTTP Error 404: Not Found

[Elaine Meng]

PDB file of 1ml1 from May 2 2017

[Eric Pettersen]

Fixed the PDB fetching. BTW, if you had fetched "from pdbe" it would have worked -- so that's another workaround until tomorrow's build.

[Eric Pettersen]

I'm not 1000% sure what I'm supposed to do here. In each chain residue 14 (a cysteine) is missing, but in the 3D structure residue 13 is clearly directly connected to residue 15 via the normal backbone atoms. The mmCIF reader puts a missing-structure pseudobond in lieu of what is pretty clearly a covalent bond, whereas the PDB reader uses a regular bond. The resultant availability of inserting a gap there for mmCIF allows the structure to be aligned to the sequence without error, but not for the PDB

The REMARK 465 records in the PDB file note the residue 14 is not located in the experiment.

[Elaine Meng]

I don’t know what is best either.  If we think it’s rare, then the workaround can be “use mmCIF instead.”  If it turns out to be a recurring type of problem, then revisit maybe trying to do something more proactive, not that I have any ideas of what that might be.  Could demote priority on this ticket and just keep it as an issue of which we are aware.

[Eric Pettersen]

The problem is that both behaviors are both right and wrong. The mmCIF behavior is right if your "believe" the SEQRES. The PDB behavior is right if you "believe" the structure.

[Elaine Meng]

That suggests trying both behaviors and then using the one that doesn’t give mismatches… perhaps not feasible though.

[Eric Pettersen]

That’s not what I’m suggesting.  I am just saying that the PDB behavior is correct (i.e. there _is_ a mismatch with the sequence) if you look at the structure.  The mmCIF behavior is correct (you need to insert a gap between those residues despite the deposited structure) if you look at the SEQRES sequence.

—Eric

[Elaine Meng]

Can we infer a possible gap from the numbering break, or is residue number always ignored?

[Eric Pettersen]

Numbering gaps don’t always imply structure/sequence gaps.  Circular permutations and fusion proteins can have numbering gaps where no sequence/structure gap exists.