AddH option to use residue templates for guidance

[Tristan Croll]

I recall this being suggested during the conversation arising from a previous ticket, but I can't seem to find it now and would like to bump it. It would be *really* useful if AddH had the option to infer chemistry from the ideal coordinates in the CCD templates from each residue rather than the model coordinates themselves. The model in #2184, for example, has quite a few carbon-carbon bonds that are slightly too short, leading to them being treated as double-bonded. I've started work on a tool in ISOLDE to find and fix residues with missing/wrong atoms, but it would be ideal if as many of these issues as possible could be avoided in the first place.

[pett]

I can imagine adding an option to use CCD templates for guidance. You are aware that CCD templates are protonated for neutrality rather than biological relevance, which changes the protonation of carboxylic acids, tetrahedral nitrogens, etc. This isn't problematic for you?

If that isn't a problem, then the other issue is that the templates don't have explicit designations of the hybridization of the atoms and that therefore I would have to hack up something to guesstimate the bond geometry from the positions of the neighboring template atoms. Certainly doable, just makes the implementation somewhat more effort than it would be if the template supplied that information.

[tic20@…]

I think the last time we discussed this, your suggestion was to use the 
templates to infer the idatm_types, and then let AddH decide on 
protonation states. That seems like the perfect approach to me.

If you were to make a temporary "real" residue from each template, then 
wouldn't the existing idatm_types framework automatically decide the 
atom types?

On 2019-07-09 21:29, ChimeraX wrote:

[pett]

I see, interesting. I would have to omit the hydrogens when making the "real" residues, since the algorithm will not assign types inconsistent with the actual protonation. Wouldn't be too hard.

[Tristan Croll]

I have a working method now to rebuild hydrogens on a given residue to match its template (and fill in missing heavy atoms into the bargain). Of course, this still has the problem of replicating the template's protonation state, which makes me wonder... would you consider adding an AddH mode which just strips and rebuilds the hydrogens on ionizable groups only? That could be *really* handy.

[pett]

I intend to implement the original plan first, then if "stripping ionizable protons" has any utility at all after that, I will give it consideration.

[Tristan Croll]

Similar problem: bad geometry on a ASN-NAG glycosidic bond can block addition of a hydrogen to the ND2 atom - see for example the /A:166 - /A:410 bond in 6j11.

[pett]

Implemented as 'template' keyword for addh (default false).

The glycosidic bond example works without templating (and fails with templating, since the template obviously does not account for a glycosidic bond).

[Tristan Croll]

That's great - will look forward to trying it!

On 2019-08-12 22:57, ChimeraX wrote: