Re: [Cdk-devel] [Cdk-user] inchi conversion changes compound

[Martin G. <gue...@po...>]

Sorry my fault. The SmilesGenerator removed the stereochemistry. Just 
noticed that I have to use .absolute() instead of .unique()

Am 03.09.2015 um 12:36 schrieb Martin Gütlein:
> Hi John,
>
> Awesome, thanks. SuppressHydrogens works good, but it suppresses Hs 
> with stereochem-information as well. Can this be avoided?
>
> > although possible you should perhaps store and index (unique check) 
> you structures separately.
> Thats what we try to not do, if possible.
>
> Martin
>
>
> Am 03.09.2015 um 11:32 schrieb John M:
>> Hi Martin,
>>
>> By design CDK separates out standardisation (H rep, tautomers, 
>> protonation) from canonicalisation (ordering). You've found the 
>> method to "sprout" hydrogens but you actually want the opposite - 
>> suppressHydrogens(mol) 
>> <http://cdk.github.io/cdk/1.5/docs/api/org/openscience/cdk/tools/manipulator/AtomContainerManipulator.html#suppressHydrogens%28org.openscience.cdk.interfaces.IAtomContainer%29>.
>>
>> I think you might be conflating two things, although possible you 
>> should perhaps store and index (unique check) you structures 
>> separately. For your example you could assign a unique tautomer but 
>> you'll be back at square one with your first example.
>>
>> O=C[CH]1Cc2[nH]cnc2CC1 -> 1,3 proton shift -> C(=O)[CH]1Cc2c(CC1)[nH]cn2
>> CC(=O)N -> 1,3 proton shift -> CC(O)=N
>>
>> Thanks,
>>
>>
>> Regards,
>> John W May
>> joh...@gm... <mailto:joh...@gm...>
>>
>> On 3 September 2015 at 09:27, Martin Gütlein <gue...@un... 
>> <mailto:gue...@un...>> wrote:
>>
>>     One more thing:
>>     I noticed that unique SMILES differentiate explicit and implicit
>>     Hydrogens, e.g. "[H]Cl" is different form "Cl". This can be
>>     solved by running
>>     AtomContainerManipulator.convertImplicitToExplicitHydrogens(mol).
>>     However, I do not like having my all Hs defined explicitly. Is
>>     there an option in the CDK to convert explict Hs back to
>>     implicit, leaving only thoses Hs as explict that are relevant?
>>
>>     Martin
>>
>>
>>     Am 03.09.2015 um 09:48 schrieb Martin Gütlein:
>>>     Hi John,
>>>
>>>     thanks for your reply, I tried to use unique (kekulized) SMILES
>>>     instead of InChIs.
>>>     Whats good is that the structure for (most) compounds is stored
>>>     correctly (i.e., I can create an IAtomContainer that is
>>>     apparently equal).
>>>
>>>     However, I found an example were the unique SMILES of two
>>>     identical structures is different (see below).
>>>
>>>     Kind regards,
>>>     Martin
>>>
>>>
>>>         for (String smi : new String[] { "O=C[CH]1Cc2[nH]cnc2CC1",
>>>     "C(=O)[CH]1Cc2c(CC1)[nH]cn2" })
>>>             {
>>>                 IAtomContainer mol = new
>>>     SmilesParser(SilentChemObjectBuilder.getInstance()).parseSmiles(smi);
>>>     System.out.println(SmilesGenerator.unique().create(mol));
>>>             }
>>>
>>>
>>>
>>>
>>>
>>>
>>>     Am 02.09.2015 um 20:58 schrieb John M:
>>>>     Just to add on - if you really want to use InChI (don't) then
>>>>     you could store the AuxInfo but the CDK doesn't have a
>>>>     conversion method that accepts it when turning it back into an
>>>>     AtomContainer.
>>>>
>>>>     I also notice you're using unique SMILES (default by old APIs),
>>>>     you probably want isomeric that a non-canonical but store
>>>>     stereochemistry.
>>>>
>>>>     IAtomContainer mol = SmilesGenerator.isomeric().create(container);
>>>>
>>>>     John
>>>>
>>>>     Regards,
>>>>     John W May
>>>>     joh...@gm... <mailto:joh...@gm...>
>>>>
>>>>     On 2 September 2015 at 19:54, John M
>>>>     <joh...@gm...> wrote:
>>>>
>>>>         Hi Martin,
>>>>
>>>>         The InChI is an identifier and not a structure
>>>>         representation it should never be used as such. For maximum
>>>>         preservation you should store compounds as Kekulé SMILES or
>>>>         Molfile. You can store additional data such as coordinates
>>>>         supplementary to the SMILES.
>>>>
>>>>         You might find a recent presentation by Noel (O Babel) and
>>>>         Rajarshi (CDK) useful:
>>>>         http://baoilleach.blogspot.co.uk/2015/08/the-whole-of-cheminformatics-best.html
>>>>
>>>>         John
>>>>
>>>>
>>>
>>
>>     -- 
>>     Dr. Martin Gütlein
>>     Phone:
>>     +49 (0)6131 39 23336 <tel:%2B49%20%280%296131%2039%2023336>  (office)
>>     +49 (0)177 623 9499 <tel:%2B49%20%280%29177%20623%209499>  (mobile)
>>     Email:
>>     gue...@un... <mailto:gue...@un...>
>>
>>
>
> -- 
> Dr. Martin Gütlein
> Phone:
> +49 (0)6131 39 23336 (office)
> +49 (0)177 623 9499 (mobile)
> Email:
> gue...@un...
>

-- 
Dr. Martin Gütlein
Phone:
+49 (0)6131 39 23336 (office)
+49 (0)177 623 9499 (mobile)
Email:
gue...@un...