_deduce_PDB_atom_name problem

[acellera]

_deduce_PDB_atom_name problem

[stefdoerr]

Hm, could you provide me with an example molecule? I got a bit confused by the explanation.

[ungsik]

Hi, I am trying the 1kqn_5 of the sc-PDB data. Sincerely, Ungsik

…

On Thu, Aug 29, 2019 at 5:43 PM Stefan Doerr ***@***.***> wrote: Hm, could you provide me with an example molecule? I got a bit confused by the explanation. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#25?email_source=notifications&email_token=AEKEUSKLTUFNXB44ERMEJRTQG6D3TA5CNFSM4IR6ZHVKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5NXRGQ#issuecomment-526088346>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEKEUSKBELCMTOVDFDGM2P3QG6D3TANCNFSM4IR6ZHVA> .

-- Ungsik Yu

[stefdoerr]

Great! I just had to add Xe to the ions list, so now it formats it correctly as in the PDB file 'XE ' instead of ' XE '. I will make a fix release soon.

[stefdoerr]

ea0bf30 Done.

Once (and if) this passes https://travis-ci.org/Acellera/moleculekit/builds/578267436 you will have a new 0.1.23 release of moleculekit with the fix.

Thanks for the report :)

[ungsik]

Hi, There are other kinds of ions not included in the ions list. Maybe changing the _deduce_PDB_atom_name function is more general. The last return statement of the _deduce_PDB_atom_name function: return ' {:<3}'.format(name) ===> return '{:<4}'.format(name) Sincerely, Ungsik

…

On Thu, Aug 29, 2019 at 6:45 PM Stefan Doerr ***@***.***> wrote: Closed #25 <#25>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#25?email_source=notifications&email_token=AEKEUSPPCCX52KCHENVIC4TQG6LCJA5CNFSM4IR6ZHVKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOTKJYZXY#event-2593361119>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEKEUSI6AZRAW7YVUILWSULQG6LCJANCNFSM4IR6ZHVA> .

-- Ungsik Yu

[ungsik]

Hi, There are other kinds of ions not included in the ions list. Maybe changing the _deduce_PDB_atom_name function is more general. The last return statement of the _deduce_PDB_atom_name function: return ' {:<3}'.format(name) ===> return '{:<4}'.format(name) Sincerely, Ungsik

…

On Thu, Aug 29, 2019 at 6:40 PM Stefan Doerr ***@***.***> wrote: Great! I just had to add Xe to the ions list, so now it formats it correctly as in the PDB file 'XE ' instead of ' XE '. I will make a fix release soon. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#25?email_source=notifications&email_token=AEKEUSJEAG2GOBNPUO35EZDQG6KQBA5CNFSM4IR6ZHVKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5N4UCY#issuecomment-526109195>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEKEUSIJFBZ67YIB5GSKNZ3QG6KQBANCNFSM4IR6ZHVA> .

-- Ungsik Yu

[stefdoerr]

No, that would break the PDB format. I would not change the string formatting. But I will try now to find all "ions" of scPDB and I'll report back.

[stefdoerr]

To make it more clear, PDB has a weird way of differentiating between ions and non-ions by the spaces before and after the atom name. For example a carbon-alpha atom will be ' CA ' while a calcium atom will be 'CA '. So you need to be very careful about where you put spaces. The good thing is that if a mol2 file of scPDB fails to read I can check in the PDB database and see how they used spaces for that atom and reproduce that behaviour.

[ungsik]

Thank you for the explanation. The attached file is generated against all files of the 2017 version of sc-PDB. The file list all lines containing Du element. Best wishes, Ungsik

…

On Thu, Aug 29, 2019 at 7:08 PM Stefan Doerr ***@***.***> wrote: To make it more clear, PDB has a wierd way of differentiating between ions and non-ions by the spaces before and after the atom name. For example a carbon-alpha atom will be ' CA ' while a calcium atom will be 'CA '. So you need to be very careful about where you put spaces. The good thing is that if a mol2 file of scPDB fails to read I can check in the PDB database and see how they used spaces for that atom and reproduce that behaviour. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#25?email_source=notifications&email_token=AEKEUSOKAU3OQRMWJ6YCU43QG6NYNA5CNFSM4IR6ZHVKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5N7CRI#issuecomment-526119237>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEKEUSMRXKV7QU2WVHCJF43QG6NYNANCNFSM4IR6ZHVA> .

-- Ungsik Yu

[stefdoerr]

Sorry for half-assing it before. Now I added all the elements of scPDB (Xe, Rb, Lu, Ga) to the element-guessing code 8472d24

This is the 0.1.24 release build which will support them all https://travis-ci.org/Acellera/moleculekit/builds/578404945

[stefdoerr]

Okay I'll reopen this issue open because fixing the reading of the unsupported metals does not really fix the guessing. Now all files are read fine but for example cadmium "CD" in 4b7b is guessed as element carbon "C". This is a quite hard problem though considering atom type names. Maybe the Du element could be used if it's a mol2 standard but adding support just for scPDB mol2 files seems like a bad hack.

[stefdoerr]

Note that you can also read the mol2 files with the SmallMol reader which marks the unknown elements as * instead of trying to guess them and then convert to Molecule if that serves you better for now.

from moleculekit.smallmol.smallmol import SmallMol
sm = SmallMol('./4b7b_1/protein.mol2')
mol = sm.toMolecule()

[stefdoerr]

Indeed Du atom types in mol2 are defined in Corina/SYBYL and TriPos. So I added some code to correctly mark them as ion/metals so that they don't get confused with carbons etc.
d8694a9

This should close the issue well.