Adrian Schreyer reported a bug in the RECAP functions connected with sanitization of charged species: https://sourceforge.net/tracker/?func=detail&atid=814650&aid=1881803&group_id=160139 This is actually the result of two problems: 1) charges aren't being properly propagated in reactions (now it's own bug: https://sourceforge.net/tracker/index.php?func=detail&aid=1882749&group_id=160139&atid=814650) 2) the Recap definitions probably shouldn't accept charged species anyway.
After a quick chat with Adrian (transcript below), I'm going to update the Recap reactions so that they no longer accept charged species. (I'll also fix the reaction bug, of course) If the "no charged atoms" rule for Recap sounds incorrect to anyone, please pipe up. -greg me: Hi Adrian Adrian: Hi Greg how can I help? me: The problem with the Recap reactions is that the + charge on the N is getting lost from reactants -> products. This is a bug in the reaction handling. (a bug which I will fix) 5:07 PM however, I think the recap rules actually shouldn't be matching the charged N in the first place so O=C[NH+](C)(C) isn't really an amide, I don't think 5:08 PM Adrian: good question 5:09 PM Now I see where the problem is 5:10 PM me: so I think I'm going to modify the default Recap rules to not allow breaking at charged atoms. Does this sound ok to you? 5:12 PM Adrian: Sounds reasonable. I probably have to modify the molecules I'm working with in terms of protonation states 5:13 PM me: since the recap reactions can be provided at run time, you will always be able to go back to the rules that accept positive species (which will be handled correctly after I fix the reaction bug) 5:14 PM Adrian: Okay, for me the canonicalization is more important 5:15 PM me: That's checked in now. You're subscribed to rdkit-devel, right? Adrian: yes me: cool. Have you seen canonicalization problems? 5:17 PM Adrian: just with the cyclic Nitrogens and the charged ones but those are already fixed or in the bug tracker... I am only struggling with importing SIMLES strings from other programs but that is not a RDKit problem 5:18 PM me: it is rather more strict than most other packages. Which is a good and bad thing Do you object to me posting the RECAP-related part of this chat to the discuss list so that it's archived? 5:19 PM Adrian: no, of course not, go ahead!
