Hi Veronica,
First a remark: the pdb is a repository and if you insist, they will accept anything you send them so, yes you can deposit this.
However, your question is off course, do I want to deposit this? One does not want to end up in the twilight gallery or deposit a structure where other crystallographers make fun of.
What I would do is to keep all residues that have some density, even at low level. Here you could go down as low as 0.6 to 0.8 sigma. However, if no density is present at all at these low levels, or if the density is scattered and unrelated to the model, I would delete it. I would only keep a few orphaned residues if they are either linked to the visible part of the protein, e.g. via a disulfide link, or have convincing interactions (Hbonds, salt links etc.) with the visible part of the protein.
If you want to keep the residues, you need help from your free Rfactor. You need to refine as good as possible the complete structure and also the structure with the weak/non-existent parts deleted. If the complete model has a lower Rfree than the truncated model, you might convince the referees that the complete model is better. Otherwise I would stick to the truncated model.
Since you mention that the disorder only occurs when a ligand is bound, it might have a (biological) function, which would provide an incentive to delete the poorly defined regions to illustrate this. However, if you obtained your ligand complex by soaking, the disorder might be caused by the fact that an interesting conformational change is blocked by the crystal packing. In this case you should try to get the complex by cocrystallization as well, otherwise you might have to do this when you receive the referee reports of your paper.
Best,
Herman
-----Ursprüngliche Nachricht-----
Von: CCP4 bulletin board [mailto:[log in to unmask]] Im Auftrag von V F
Gesendet: Mittwoch, 11. Januar 2017 12:25
An: [log in to unmask]
Betreff: Re: [ccp4bb] To delete main chain or not
Hi all,
Since some emails were not REPLY ALL, I have answered them here:
The structure solved with MR; high sequence identity. Disorder occurs only when ligand is bound. Model is almost 90 % complete except the disordered region. Please let me know if I can deposit structure as it is. I took example 1fcc (also low resolution). Seems many residues are outside Ramachandran plot (even Residue 266 has no density at all).
Does it mean I can deposit this?
> Yes, it is okay to exclude few residues or even short stretches of loops/helices/strands if there is no reliable/convincing electron density for them even at low map contours because if there is no convincing density you cannot reliably model the local structure.
>
> However, if you refine with tight NCS, which you probably should do at this resolution unless there are significant differences in the 4 molecules, you can also choose to keep residues for which there is no reliable density in one of the molecules, provided those residues are well resolved in the other molecules and refine well (density, B-factors). Also use TLS refinement.
>
> Irrespective of which of the 2 options you choose, make sure to add a remark in the PDB header about the option chosen, so that others know what was done and if you write it up in a manuscript, describe it there as well.
> What is your R/Rfree currently? And what is the CC(1/2) of your data
> at 2.8
> If you haven’t done so already, I’d delete suspect, refine and see if difference density comes back for it.
The density looks similar to it.
> Perhaps the domain is disordered and really needs to be deleted.
> Other possibilities are that it is in an unexpected orientation or conformation.
> You may need several iterations of deleting, rebuilding and adapting.
>
> Were the input phases experimental or from a molecular replacement model, and if the latter how complete was the model?
>
>
|