JW – Generally want to make our interfaces with OpenEye and RDKit more reliable and better documented. This would take the form of providing clear documentation of the expectations of each (implicit Hs? Aromaticity? Mol fragments/radicals/dangling bonds? Metadata?)
I think the best start to this work would be to find a better molecule test set than MiniDrugBank.sdf, round trip the molecules through both OE+RDK, and check for differences and whether they’re significant.
Also, if there’s some way to determine whether the SMARTS matching could be different for the different toolkits
Final notes on this would live in method docstrings and the developer documentation