No results for ~22,000 SMILES with release 2.0.1

[redaschi]

Hello, last December I found Dimorphite-DL when looking for an alternative to the software we are currently using to get the protomer of a compound at a given pH. Having no Python experience, I finally managed to install it via https://github.com/conda-forge/dimorphite-dl-feedstock. My test over app. 110,000 SMILES looked promising, and this week I wanted to repeat it with newer data when I saw that the project is updated again, which is great news! I installed the latest version 2.0.1 from https://pypi.org/project/dimorphite-dl/, but now I got no results for app 22,000 SMILES. I looked at an example and tested it on the command-line to ensure the bug is not in my Python script:

dimorphite_dl --ph_min 7.0 --ph_max 7.0 --precision 0.0 "CCC(=O)O"
CCC(=O)[O-]
dimorphite_dl --ph_min 7.0 --ph_max 7.0 --precision 1.0 "C=C[C@@](C)(O)CC/C=C(\C)CCC=C(C)C"
--> No result

Then I tried the oldest version I see on pypi (1.2.5) and it works with my test example:

dimorphite_dl --silent --min_ph 7.0 --max_ph 7.0 --pka_precision 0.0 --smiles "C=C[C@@](C)(O)CC/C=C(\C)CCC=C(C)C"
C=C[C@@](C)(O)CC/C=C(\C)CCC=C(C)C

Please let me know if you need more examples or other information to investigate this.
Best regards,
Nicole

[nacefguessaymi]

Hello Nicole,
Thank you for using our tool! We have recently released an update to address this issue. This should be fixed in version 2.0.2. You should be able to fix this by installing the new version on pip.
Please let me know if the issue persists!
Thanks!

[redaschi]

Dear Nacef,
I installed 2.0.2 and have now SMILES for nearly all compounds :-) Only 5 compounds differ compared to version 1.2.5:

2 compounds still give no result - they look quite exotic to me, I add their SMILES here in case you are interested:

CHEBI:37269 O=[N+]1[O][Ce-2]23456([O][N+](=O)[O-]->2)([O][N+](=O)[O-]->3)([O][N+](=O)[O-]->4)([O][N+](=O)[O-]->5)([O][N+](=O)[O-]->6)<-[O-]1
CHEBI:193050 [2H]OC(=O)C(F)(F)F

3 compounds give a different result, I can live with that because they do not look like biological compounds:
https://www.ebi.ac.uk/chebi/beta/CHEBI:51433
https://www.ebi.ac.uk/chebi/beta/CHEBI:51434
https://www.ebi.ac.uk/chebi/beta/CHEBI:51439

Many thanks for your support and this quick fix!
Nicole

[nacefguessaymi]

Hi Nicole,
Glad to hear the update fixed this issue!
Thanks for sending all these test cases. They are really helpful in testing out edge cases for our software!
For the two compounds you sent, the two smiles strings giving no results, they are not being read as valid by dimorphite.
For CHEBI:37269 it seems the smiles string you attached is different to the one provided by Chebi which is O=N1=[O][Ce--]23456789%10(O1)ON(=O)=[O]2.O=N(O3)=[O]4.O=N(O5)=[O]6.O=N(O7)=[O]8.O=N(O9)=[O]%10.
When running this through dimorphite, it yields the result [O-][N+]1[O][Ce-2]23456([O][N+]([O-])=[O]->2)([O][N+]([O-])=[O]->3)([O][N+]([O-])=[O]->4)([O][N+]([O-])=[O]->5)([O][N+]([O-])=[O]->6)<-[O]=1.
Interestingly, under pubchem the entry is [N+](=O)(O)[O-].[N+](=O)(O)[O-].[N+](=O)(O)[O-].[N+](=O)(O)[O-].[N+](=O)(O)[O-].[N+](=O)(O)[O-].[Ce] which yields O=[N+]([O-])[O-]. It seems like the cesium is a central group for coordinating the nitriles around it through ionic interactions. I'm not sure which parts of this compound is relevant to your work.
CHEBI:193050 on the other hand it seems like the [2H] in the string provided is a deuterium. This is being read as two hydrogens attached to the oxygen, which is more than the allowed bonds. I think theoretically this should function as a simple hydrogen, so using the string [H]OC(=O)C(F)(F)F yields O=C([O-])C(F)(F)F.
For the other three compounds, like you mentioned they do not really look like biological compounds, and they are pretty large. I believe the cause for the difference in results is there was a bug where each site was being protonated individually for each structure given, rather than all sites being protonated in one sturcture when using a pKa precision of 0.0. I think this may be the reason the difference in your case, but I cannot know for sure. If you would like me to look into this further the exact command that you ran for these compounds and the results (previous and current) would be need to source this difference.
Please let me know if there are any issues as well!