invalid SMILES in `canon_building_block_mol2idx_no_isotope.json`

[newalexander]

Thanks for making this work accessible. When I've tried to use this code, I sometimes run into issues during desp.search.

File "script.py", line 23, in <module> result, route, searcher = desp.search( File ".../desp/desp/desp/DESP.py", line 135, in search result = searcher.run_search() 
File ".../desp/desp/search/desp_search.py", line 231, in run_search _ = self.expand_fwd(best_node, self.search_graph) 
File ".../desp/desp/search/desp_search.py", line 119, in expand_fwd predictions = self.fwd_model.predict( 
File ".../desp/desp/inference/forward_predictor.py", line 171, in predict output = run_bimolecular_reaction( 
File ".../desp/desp/inference/utils.py", line 149, in run_bimolecular_reaction reactants = rdchiralReactants(".".join(reactants))
File ".../desp-env/lib/python3.10/site-packages/rdchiral/initialization.py", line 101, in __init__ self.reactants = initialize_reactants_from_smiles(reactant_smiles) 
File ".../desp-env/lib/python3.10/site-packages/rdchiral/initialization.py", line 181, in initialize_reactants_from_smiles 
Chem.AssignStereochemistry(reactants, flagPossibleStereoCenters=True) 
Boost.Python.ArgumentError: Python argument types in rdkit.Chem.rdmolops.AssignStereochemistry(NoneType) did not match 
C++ signature: AssignStereochemistry(RDKit::ROMol {lvalue} mol, bool cleanIt=False, bool force=False, bool flagPossibleStereoCenters=False)

I believe the cause is that a small number of the SMILES keys in canon_building_block_mol2idx_no_isotope.json are invalid, making Chem.MolFromSmiles and similar functions return None.

['[IH2-2]',
 'CCN(CC)c1ccc2c(-c3ccccc3C(=O)NCCCC(NC(=O)COCCOCCN)C(=O)OC(C)(C)C)c3ccc(=[N+2](CC)CC)cc-3oc2c1',
 'F[P+](F)(F)(F)(F)F',
 'CCc1nc(SCC(=O)NC2C(=O)N(c3ccccc3)[N+2](C)=C2C)[nH]c(=O)c1C#N',
 'COc1ccc(Nc2nc(N/N=C/C3=[C+]4=CC=CC=[C+]4=C[C+]4=CC=CC=[C+]=43)nc(Nc3ccccc3)n2)cc1',
 'O=C1NC2=[N+2](C=C(c3ccccc3)S2)c2sc3c(c21)CCCC3',
 'F[B-2](F)(F)C[NH+]1CCCCC1',
 'CCOC(=O)C1=C(CSC2=[N+2](C)C=CN2)NC(=O)NC1c1ccc(C)cc1',
 'CC(C)c1ccc(/C=C2/S/C(=N\\C3=[C+](=O)C=CC=[C+]3=O)N(c3ccccc3)C2=O)cc1',
 'CC(c1cccc(C(=O)c2ccccc2)c1)[C+](=O)OC[C@@H](O)[C@@H](O)[C@H](O)[C@@H](O)CO',
 'Cc1ncsc1CCO[P](=O)(=O)O',
 'CC(C)(C)OC(=O)N1CCN(C[B-2](F)(F)F)CC1',
 'CN(C)/C=C/C=[N+2](C)C',
 'O=[N+]([O-])c1cccc(C2=CC=[N+2]([O-])C=C2)c1',
 'O=S1(=O)C[C@@H]2SC3=[N+2](C=CC=C3)[C@@H]2C1',
 'CC1=[N+2](C)N(c2ccccc2)C(=O)C1NC(=O)Cn1nnc2sc3c(c2c1=O)CCC(C(C)(C)C)C3',
 'C[C@H](C1=NC=NC=[C+]1F)C(O)(Cn1cncn1)c1ccc(F)cc1F',
 'CC1=[N+2](C)N(c2ccccc2)C(=O)C1NC(=O)Cn1nnc2sc(-c3ccccc3)cc2c1=O',
 'C[P+](C)(C)(C)(C)C',
 'F[Si](F)(F)(F)(F)F',
 '[AlH3+3]',
 'F[B-2](F)(F)CCl',
 '[H]C([H])([H])C(=O)N[C@@H](CSC(CO)[C+]1=CC=CC=C1)C(=O)O',
 'C[N+]12CCc3cc4c(c(O)c3C1C[C+]1=C(C2)C2=C(C=C1)OCO2)OCO4',
 'CCCC[P+](c1ccccc1)(c1ccccc1)[C+]1=CC=CC=C1',
 'COC[B-2](F)(F)F',
 'F[PH](F)(F)(F)(F)F',
 'F[B-2](F)(F)CC1CCC1',
 'F[B-2](F)(F)CC1CCCC1',
 'CC(=O)N[C@@H](CSC(CO)[C+]1=CC=CC=C1)C(=O)O',
 'O=C(Nc1cnccn1)[N+2]1=NNc2ccccc21',
 'CC1=[N+2](C)N(c2ccccc2)C(=O)C1NC(=O)Cn1nnc2sc3c(c2c1=O)CCC(C)C3',
 'CC(C)C1=C(C(=O)Nc2ccccc2)C(c2ccccc2)=[C+](c2ccc(F)cc2)[N+]1([O-])CC[C@@H](O)C[C@@H](O)CC(=O)O',
 'F[P](F)(F)(F)(F)F',
 'N#CCC[B-2](F)(F)F',
 'F[B-2](F)(F)CCCc1ccccc1',
 'Cc1ccc(S(=O)(=O)NC[B-2](F)(F)F)cc1',
 'CC(C)(C)OC(=O)NC[B-3](F)(F)F',
 'Fc1ccc([B-2](F)(F)F)cc1',
 'CC(C)(C)OC(=O)N1CCC(OC[B-2](F)(F)F)CC1',
 'CC1=[N+2](C)N(c2ccccc2)C(=O)C1NC(=O)Cn1nnc2sc3c(c2c1=O)CCCC3',
 'O=C(Nc1ccccn1)[N+2]1=NNc2ccccc21',
 'O[C@@H]1C[C@H]2CC[C@@H](C1)[C+]21CCCC1',
 'Oc1ccc([B-2](F)(F)F)cc1',
 'Cc1ccc([N-]C(=O)[C@H]2CCC[N+5]2Cc2ccccc2)c(/C(=[N+5]/CC(=O)[O-])c2ccccc2)c1']

I don't know if these were incorrectly constructed or if something else went wrong during the preprocessing.