Depends on the data, journal reactions are more esoteric. People like to try and eval/test atom mapping with hard/rare reactions. Common reaction mechanisms used day-to-day are common because they're useful building blocks :-).
Something I always thought it (IBM RxnMapper) would benefit from is some basic checks, it actually gives multiple results with a probability. But some basic empirical calculations could be used to clean things up - e.g. number of C-C bonds broken, bonds formed etc. At NextMove we have a "amap bench" utility (I think from 2013) which tells us how good a mapping is. I still need to integrate this to remove some dubious ones from Indigo but my view is supplementing the RXNMapper with this would be good. https://nextmovesoftware.com/posters/Sayle_HazELNutExtractReactionsELN_ACS_201309.pdf https://nextmovesoftware.com/talks/Sayle_ReactionProcessing_Sheffield_201307.pdf On Thu, 20 Oct 2022 at 10:52, Uli Fechner <u...@pending.ai> wrote: > Okay, thank you. > > We do use RXNMapper and in our experience the numbers they state in their > publication are a bit optimistic... > > On Thu, Oct 20, 2022 at 7:25 PM John Mayfield <john.wilkinson...@gmail.com> > wrote: > >> It does not, personally today I would use the IBM AI based one. >> >> On Thu, 20 Oct 2022 at 07:05, Uli Fechner <u...@pending.ai> wrote: >> >>> Hi, >>> >>> I vaguely remember that there is/was code in the CDK for atom-to-atom >>> mapping of reactions (wasn't that RDTool?). However, I cannot find that now. >>> >>> RDTool by Asad uses CDK. But its maintenance seems to be lacking a bit >>> (last pre-release based on CDK 2.5 is back from Mar 2021). >>> >>> Does CDK provide atom-to-atom mapping for reactions at the moment? Are >>> there any open-source tools out there other than RDTool that offer that >>> functionality and use CDK as a library? >>> >>> Any help is appreciated. >>> Uli >>> >>> _______________________________________________ >>> Cdk-user mailing list >>> Cdk-user@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/cdk-user >>> >>
_______________________________________________ Cdk-user mailing list Cdk-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/cdk-user
