Voxelization of PDB 1E2X ends strange error message

[acellera]

Voxelization of PDB 1E2X ends strange error message

[stefdoerr]

Ugh, this is a protein preparation / bonding issue. pdb2pqr adds terminal atoms to GLN 227 during the protein preparation but the distances are probably a bit off for atom 3514 so the bonds are not guessed correctly for it and it's not considered "protein" anymore.

In [8]: prot = prepareProteinForAtomtyping(prot)

In [11]: prot.atomselect('not protein', indexes=True)                                                                                                                                                              
Out[11]: array([3514], dtype=int32)

In [12]: prot.get('resname', sel='not protein')                                                                                                                                                                    
Out[12]: array(['GLN'], dtype=object)

In [13]: prot.get('resid', sel='not protein')                                                                                                                                                                      
Out[13]: array([227])

If that atom is not close to any site of interest I would suggest removing it with prot.remove('index 3514') and continuing with the voxelization. A bit of an extreme solution but sometimes you really can't win with such complex protocols and external software.
If it occurs in many other cases feel free to report them and I will check if anything can be done about it.

[keisuke-yanagisawa]

I catch the reason. Hmmmmm...
Actually I met the error several times, and thus far I avoided those by downloading atomtyper.py and voxeldescriptors.py and modified notallowed = ~(protsel | metalsel) to notallowed = ~(protsel | metalsel | stdaasel) (there is an another reason to decide modify atomtyper.py: some protein has PB, BA, and other metals, btw).

Actually, I utilized preprocessed PDB files, thus I could not guarantee which PDB record raise the error in original state. If you want, I will check it for original PDB files of interest.

[keisuke-yanagisawa]

But my second thought is that below code is probably more robust:

prot = Molecule("pdb1e2x.ent.gz")
prot.filter("protein or ions") # remove H2O and SO4
prot = prepareProteinForAtomtyping(prot)
prot.filter("protein or ions") # remove isolated atoms

prot_vox, prot_centers, prot_N = getVoxelDescriptors(prot, buffer=1)

Is it valid way to avoid the issue for you?

[stefdoerr]

Yes but maybe take a look at how many atoms get removed. the filter method returns the atoms which were removed so check that it doesnt suddenly delete 200 atoms or something ;)

…

On Tue, Oct 15, 2019 at 6:50 PM K. Yanagisawa ***@***.***> wrote: But my second thought is that below code is probably more robust: prot = Molecule("pdb1e2x.ent.gz") prot.filter("protein or ions") # remove H2O and SO4 prot = prepareProteinForAtomtyping(prot) prot.filter("protein or ions") # remove isolated atoms prot_vox, prot_centers, prot_N = getVoxelDescriptors(prot, buffer=1) Is it valid way to avoid the issue for you? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#36?email_source=notifications&email_token=AB4RLAV2ZZQCZQ63ZKLECMLQOXRC5A5CNFSM4JARSV52YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBJIVJQ#issuecomment-542280358>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB4RLAVD3FA44GU2V4IJGVLQOXRC5ANCNFSM4JARSV5Q> .

[keisuke-yanagisawa]

Okay I understand! I will monitor the numbers.

[stefdoerr]

I will close the issue for now. If you have any further questions feel free to comment here or open a new one