Warning messages when trying to load_bbbp

#1

Hello all, :wave:

I am new to DeepChem and I was going over the notebooks tutorials where I encountered few warning messages that I am not sure whether it is okay to ignore them and continue or should I do something else to fix the reasons behind them.

The warning messages resulted from running the code lines below

tasks, datasets, transformers = dc.molnet.load_bbbp()
train_dataset, valid_dataset, test_dataset = datasets

The warning messages are too long to be included here, so I am just going to include a sample from each type below:
[12:58:49] Explicit valence for atom # 1 N, 4, is greater than permitted
WARNING:deepchem.feat.base_classes:Failed to featurize datapoint 59, None. Appending empty array
WARNING:deepchem.feat.base_classes:Exception message: Python argument types in
rdkit.Chem.rdmolfiles.CanonicalRankAtoms(NoneType)
did not match C++ signature:
CanonicalRankAtoms(RDKit::ROMol mol, bool breakTies=True, bool includeChirality=True, bool includeIsotopes=True, bool includeAtomMaps=True)
[12:58:49] WARNING: not removing hydrogen atom without neighbors

I would appreciate your feedback on that.

Thank you in advance.

Note: I am using Google Colab

#2

These messages are mostly expected, but it may be worth sorting them out if you are running on molecules you care a lot about. They typically mean something is wrong in the molecule (which is common in online SMILES databases) and rdkit is taking some non-standard action as a result.

#3

Thank you for your reply.

To clarify, are you suggesting that I identify the molecules causing these messages and consider removing them from the dataset? If so, could you recommend the best approach or tools within DeepChem or RDKit to efficiently identify and handle these problematic molecules?

Thank you again for your help!