RDKit chirality error causing failure in 144 systems of test set

[MatthewMasters]

trafficstars

I'm trying to reproduce the results from the paper and so after generating the ESM embeddings I entered: python -m inference --protein_ligand_csv data/testset_csv.csv --out_dir results/user_predictions_testset --inference_steps 20 --samples_per_complex 40 --batch_size 10 --actual_steps 18 --no_final_step_noise

However, this failed for 144 systems in the test set with an error like:

Failed on ['data/PDBBind_processed/6d08/6d08_protein_processed.pdb____data/PDBBind_processed/6d08/6d08_ligand.sdf']
Invariant Violation
	no eligible neighbors for chiral center
	Violation occurred on line 239 in file Code/GraphMol/FileParsers/MolFileStereochem.cpp
	Failed Expression: nbrScores.size()
	RDKIT: 2022.03.5
	BOOST: 1_75

which occurs during w.write(mol) at datasets/process_mols.py:487.

I was able to circumvent the issue by saving as PDB instead but since others have been able to reproduce the results, I'd like to know what the issue is. I've tried multiple versions of RDKit (2022.03.5 and 2022.09.3).

[cisprague]

@MatthewMasters, can you elaborate on how you circumvented this issue by saving it as PDB instead? I am having the same issue. Thanks!

[Geraldene]

@cisprague where you able to resolve this issue?

[Neo-Flash]

I solved this problem by modifying the "def write_mol_with_coords" function.