Re: [Rdkit-discuss] Query on a failed molecule from SureChEMBL

[Nicolas B. <nb...@eb...>]

Hi Lewis,

Currently structures are generated automatically in SureChEMBL so this kind of error unfortunately happens…

My colleagues will address this issue as soon as possible.

Cheers,
Nicolas
-----------------------------------------------
Dr Nicolas Bosc
Data Mining and Analysis Scientist
ChEMBL group
EMBL-EBI
Wellcome Genome Campus
Hinxton, Cambridge, CB10 1SD
United Kingdom

nb...@eb...
+44 1223 492519


> On 15 Dec 2021, at 20:42, Lewis Martin <lew...@gm...> wrote:
> 
> Thanks a lot Greg! That is indeed very helpful. 
> 
> Just to know that the molecule is odd is helpful too. The mol blocks appear to be V2000 format and have names like "Mrv0541 03021215572D" which says ChemAxon Marvin to me, but I'm still unsure why SureChEMBL would use such a representation (it doesn't look like a faithful transcription from the source patent). Off-topic, but if anyone happens to have an insight or connection with SureChEMBL, please do reach out!
> 
> Cheers
> Lewis
> 
>  
> 
> 
> On Wed, Dec 15, 2021 at 4:24 PM Greg Landrum <gre...@gm... <mailto:gre...@gm...>> wrote:
> Hi Lewis,
> 
> Dealing with all the strange chemical representations that show up "in the wild" is an ongoing struggle.
> 
> Your first example is pretty clearly intended to be an azide and we can certainly add a rule to normalize that one to what the RDKit expects it to be (there already is a rule for C-N=N#N, but that doesn't help here.). That won't happen before the next feature release though.
> 
> I'm not really sure what the intent was for the two four-coordinate neutral Ns in the second molecule, so I think it's unlikely that we'd add a standard cleanup for one.
> 
> However! The good news is that there's a pretty easy (and efficient) way to fix this yourself. We added a new method to chemical reactions in the 2021.09 release which allows you to modify a molecule in place (subject to some constraints). This is ideal for doing cleanup transformations like these.
> 
> This gist shows how to write reaction rules for your cases (I guessed for what the Ns are supposed to be) and then use them:
> https://gist.github.com/greglandrum/8fd229bc6bf6c734d1c21da7f2bebebb <https://gist.github.com/greglandrum/8fd229bc6bf6c734d1c21da7f2bebebb>
> 
> Hope this helps,
> -greg
> 
> 
> On Wed, Dec 15, 2021 at 12:21 AM Lewis Martin <lew...@gm... <mailto:lew...@gm...>> wrote:
> Hi All, 
> Reading molecules from a bulk download of SureChEMBL, I come across a fair few molecules that fail to parse. Not sure whether they SHOULD parse or not. 
> 
> Here is an example: https://www.surechembl.org/chemical/SCHEMBL386 <https://www.surechembl.org/chemical/SCHEMBL386>
> with SMILES code: COC(=O)C1=C(C=CC=C1)C1=CC=C(C[N+]#[N]=[N-])C=C1
> 
> Even reading the SMILES code one can see that there are too many bonds in there - a nitrogen triply bonded and doubly bonded to other atoms. 
> 
> Another example: https://www.surechembl.org/chemical/SCHEMBL33957 <https://www.surechembl.org/chemical/SCHEMBL33957>
> smiles: NC(N)=[NH]C1=NC(CSCC[NH]=CNS(=O)(=O)C2=CC=C(Br)C=C2)=CS1
> 
> Again, valence for a nitrogen is off. 
> 
> Should I expect to parse these with RDKit? Might there be some way around this? It's a significant fraction of the molecules in SureChEMBL. 
> 
> Thanks team!
> Lewis 
> _______________________________________________
> Rdkit-discuss mailing list
> Rdk...@li... <mailto:Rdk...@li...>
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss <https://lists.sourceforge.net/lists/listinfo/rdkit-discuss>
> _______________________________________________
> Rdkit-discuss mailing list
> Rdk...@li...
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss