rdkit update breaks entry validation #62
Description
Problem description
With rdkit==2024.3.6, the validation of all MITE entries passes [i.e. substrate(s)>reactionSMARTS>product(s)]. However, in more recent versions of rdkit, the validation for several MITE entries fails due to unknown reasons, presumably issues with chirality.
Example
reaction SMARTS: [#6:1]1(:[#6:36]:[#7:35;h1]:[#6:30]2:[#6:31]:[#6:32]:[#6:33]:[#6:34]:[#6:29]:1:2)-[#6:2]\\[#6@:3]1-[#6@:13]2-[#6@@:7]3(-[#6:27](=[#8:28])-[#6:26]=[#6:25]-[#6:24]-[#6:23]-[#6:21](-[#6:22])=[#6:20]-[#6@@:18](/[#6:19])-[#6:17]-[#6:16]=[#6:15]/[#6@:8]-3-[#6:9]=[#6:10](-[#6:14])-[#6@:11]/2/[#6:12])/[#6:5](=[#8:6])-[#7:4]-1>>[#6:1]1(:[#6:36]:[#7:35;h1]:[#6:30]2:[#6:31]:[#6:32]:[#6:33]:[#6:34]:[#6:29]:1:2)-[#6:2]\\[#6@:3]1-[#6@:13]2-[#6@@:7]3(-[#6:27](=[#8:28])-[#6:26]=[#6:25]-[#6:24]-[#6:23]-[#6:21](-[#6:22])=[#6:20]-[#6@@:18](/[#6:19])-[#6:17]-[#6:16]=[#6:15]/[#6@:8]-3-[#6@@:9]3\\[#8]-[#6@:10]-3(/[#6:14])-[#6@:11]/2/[#6:12])/[#6:5](=[#8:6])-[#7:4]-1
substrate: c1(C[C@@H]2NC(=O)[C@@]34[C@H](C=C([C@@H](C)C23)C)C=CC[C@H](C)C=C(C)CCC=CC4=O)c2c(cccc2)[nH]c1 |c:21,t:16,26|
expected product: c1(C[C@H]2C3[C@H](C)[C@]4(C)[C@@H](O4)[C@H]4[C@@]3(C(N2)=O)C(C=CCCC(C)=C[C@H](CC=C4)C)=O)c2c(cccc2)[nH]c1
error with rdkit>2024.3.6
ValueError: Reaction did not lead to all expected products.
Expected products:
CC1=C[C@@H](C)CC=C[C@H]2[C@@H]3O[C@]3(C)[C@@H](C)C3[C@H](Cc4c[nH]c5ccccc45)NC(=O)[C@]32C(=O)C=CCC1
Generated products:
CC1=C[C@@H](C)CC=C[C@H]2[C@H]3O[C@]3(C)[C@@H](C)C3[C@H](Cc4c[nH]c5ccccc45)NC(=O)[C@]32C(=O)C=CCC1
As can be seen from the expected vs generated products, the difference is a change in chirality (from @@ to @)
This is possibly related to #60 and atropisomeric bonds
Downstream problems and possible solutions
Currently, this issue is mitigated by pinning rdkit==2024.3.6; however, this restricts the MITE project to >=3.12.0,<3.13.0, which is not sustainable.
The proposed solution is to manually fix entries/reaction SMARTS that are not passing with rdkit>2024.3.6