phi-grib / data_curation Goto Github PK

Generation of a python code that handles SMILES curation semi automatically.
Implement it in a general way so the rest of the group can use it for different projects.

Check the following links:

https://github.com/greglandrum/RSC_OpenScience_Standardization_202104/blob/main/Standardization%20and%20Validation%20with%20the%20RDKit.ipynb

https://github.com/chembl/ChEMBL_Structure_Pipeline/wiki/Work-done-by-each-step

rdkit/rdkit#3998

Some molecules fail at Sanitization step. There should be a function that handles this errors and allows to classify the molecule despite this error.

Example:

from rdkit import Chem
error_struc = [Na+].[Na+].[Na+].[Na+].[Cu]1Oc2cc(ccc2N\\N=C\\3C(=O1)c4c(N)cc(cc4C=C3[S]([O-])(=O)=O)[S]([O-])(=O)=O)c5ccc6N\\N=C\\7C(=O[Cu]Oc6c5)c8c(N)cc(cc8C=C7[S]([O-])(=O)=O)[S]([O-])(=O)=O
Chem.MolFromSmiles(error_struc)
RDKit ERROR: [13:08:10] Explicit valence for atom # 16 O, 3, is greater than permitted

This either requires a specific handler for this error or a workaround to correct the molecule at make it valid for sanitization.

Some molecules such as CO are being miss-classified.
In this case, CO should be considered inorganic but it's classified as organic.
The filter applied in principle should be able to establish this difference, since first checks for molecules with C or c atoms, then runs Chem.AddHs to see if the molecule has carbons and hydrogens and if not, checks if it has halogen elements (Cl, F, Br, I).
It None is returned from those filters, the molecule is considered inorganic.

Filters need to be checked with this particular cases to see what is happening.