Handling of poor geometry in AddH

[Tristan Croll]

Background: ISOLDE's chirality restraints aren't yet perfect - they will *restrain* a chiral centre to the correct chirality, but they're unable to *flip* the chirality of an incorrect centre. They try hard, but the atoms are unable to find their way past each other. In the case of a threonine side chain for example, the CB-CG2 and CB-OG1 bonds end up almost co-planar with the HB preventing them from flipping further.

Something I'm planning to take care of (with a tool to flag incorrect chiral centres to the user and aid them in interactively correcting them), but it does flag up what I see as an issue with how AddH decides on adding hydrogens. If I remove and re-add hydrogens, then a threonine like the above is treated as a ketone (only two hydrogens added to CG2, none to OG1). I've also seen it treat one of the C-C bonds in a lysine sidechain as double-bonded, for a less-discernable reason.

It's great (and really quite impressive) that AddH is able to infer hydrogenation based on geometry alone, but I also think it needs a mode that sticks to strict dictionary definitions for each residue. Bad geometry is an unfortunate fact of life (albeit a problem that one would hope is reducing with time), and depending on the use-case the current behaviour of AddH risks making geometry errors worse rather than helping to fix them.

[Eric Pettersen]

Hi Tristan,

As far as I know, for amino acid and nucleic acid residues with PDB-standard names and PDB-standard heavy atom names, ChimeraX uses a lookup table for the atom type (which determines the protonation). Can you provide an example file where you are not getting expected protonation?

--Eric

[tic20@…]

PDB attached. There doesn't *appear* to be anything wrong with the heavy 
atom names in the offending residues (and I removed all hydrogens before 
re-running addh). If I delete all hydrogens, save a PDB, re-open it and 
re-run addh on that, the hydrogens still come out wrong.

Bad threonines: ['B: 263', 'C: 263', 'F: 123', 'G: 52', 'R: 52', 'R: 
54']
Bad lysines: ['G: 21', 'Q: 119', 'V: 93']
Bad valine: ['P: 127']


On 2018-07-05 22:44, ChimeraX wrote:

addh_problems.tar.gz​

[tic20@…]

Added by email2trac

[Tristan Croll]

Whatever is causing it, the problem isn't unique to this particular structure. Similar issues with another structure today (which I'm afraid I can't share). Using the latest Linux build of ChimeraX.

One thing both structures have in common is that each has at least one truncated sidechain. Other than that, I haven't got much.

[Eric Pettersen]

Fixed. The atom-type-template dir wasn't getting installed.