Brilliant. I'll test it out - I have a set of tautomers somewhere I've pulled out from ChEMBL. Regarding keto-enol, if you do add this in, could you put it within an option just in case it leads to weird results? I have the impression from others that this may be the case.

@baoilleach - did you have a chance to test this? I'd like to merge for 3.1 - even if it needs further improvement, it's a big step forward.

Just to follow-up (only a year too late). The new code is much improved indeed. However, I'm still able to find some test failures, e.g. S=c1[nH]c(c([nH]1)c1ccccc1)c1ccccc1 and Sc1[nH]c(c(n1)c1ccccc1)c1ccccc1 don't resolve to the same form. I can provide a longer list if @timvdm or anyone else is interested.

A separate issue is that it's not possible to use the functor from the Python API. I know that OEChem have solved this by somehow enabling Python functions to be passed in. An alternative would be to do as Tim did previously for automorphisms (I think), where there is an API function that either has a default functor or acts as a convenience function to just return a vector<OBMol*> of results (perhaps with an optional maxsize argument?).

@baoilleach This issue is fixed in #2410

A longer list of examples is welcome if there are still issues. I'll have a look at the callback issue when I have time.

Great stuff @timvdm. I'll pull a list together.

Related to this I've been searching for other implementations which may be of interest. Roger's one is hidden behind his slides: https://www.daylight.com/meetings/emug99/Delany/tautomers/. Meanwhile, @johnmay has an implementation discussed in his thesis available on a CDK branch: https://github.com/cdk/cdk/tree/sd-tautomer/tool/tautomer/src/main/java/org/openscience/cdk/tautomer.

Something to note is that when it comes to the step of trying to assign double/single bonds to the system, it should be possible to call the OBKekulize function with a properly prepared OBMol. This function can kekulize any system that's possible to kekulize and returns 'false' otherwise.

Happy to send a few recent papers:
https://pubs.acs.org/doi/10.1021/acs.jcim.9b01156 => https://cactus.nci.nih.gov/download/tautomer/
https://pubs.acs.org/doi/10.1021/acs.jcim.9b01080 (includes SMIRKS)

...the second of which has a contribution from me. :-)

Oh, I hadn't noticed the "O'Boyle and Sayle" spreadsheet. 😉

My group has a longer-term project to use QM calc to generate more tautomer preferences, but that's probably several months to a year out.

Hi @timvdm, apart from a logical error in my own code (which I need to think about), the OB code still had trouble with the following pairs:

*=Nc1[nH]cc2-c(c1)c1ccccc1n2,*=Nc1ncc2c(c1)c1ccccc1[nH]2
*=Nc1ccc(cc1)NC(Nc1ccccc1)=S,*=Nc1ccc(cc1)N=C(Nc1ccccc1)S
*=Nc1cccc(c1)NC(N*)=S,*=Nc1cccc(c1)NC(=N*)S
*=Nc1ccc(cc1)NC(=N*)S,*=Nc1ccc(cc1)NC(N*)=S

It also had trouble with protonated nitrogens, which hopefully is a simple atom-typing issue:

*N=CC(c1[nH]ncn1)=CN=[N+](c1ccccc1)*,*N=CC(c1n[nH]cn1)=CN=[N+](c1ccccc1)*
*NC(c1ccc[n+](c1)*)=O,*N=C(c1ccc[n+](c1)*)O