[buster-discuss] different residue alternate conformer

[Clemens Vonrhein]

Dear Ed.

On Mon, Feb 04, 2019 at 04:21:54PM -0500, Edwin Pozharski wrote:
> For some reason, I need to model a structure with alternate conformer in
> certain position that represents different amino acids.

A case of micro-heterogeneity, yes.

> As I expected, buster throws an error at pdbcheck step.

It should be able to handle this after a bit of TLC - but you need to
work with a different residue identifier (chain/residue number) and an
adequate LINK record.

> I can force the issue by disabling various pdbchecks and
> gellychecks,

Yes ... but these are usually there to highlight some potential issues
with the input files, like missing/wrong LINK records. So we usually
don't recommend switching these off.

> but what happens is that the second conformation gets renamed to
> match the amino acid name.

Yes - you need something like this (taking 5ZA2 as an example with B64
modelled as SER and SEP residues):

 (1) Ensure each residue has a different altConf identifier (here:
     SEP=A and SER=B): this is so that these two residues don't see
     each other (which would result in bad contacts).

     Alternatively you could also create an explicit EXCLUDE card

       EXCLUDE A|101 B|202

     in a file that you pass to BUSTER via 'refine -Gelly your.file
     ...'.

 (2) Give one of these residues a different residue number,
     e.g. adding 1000 to it. For 5ZA2 this would mean going e.g. from

       HETATM 3150  N  ASEP B  64      26.196   6.180   0.656  0.50 19.35           N  
       HETATM 3151  CA ASEP B  64      25.897   6.120  -0.774  0.50 18.88           C  
       ...

     to

       HETATM 3150  N  ASEP B1064      26.196   6.180   0.656  0.50 19.35           N  
       HETATM 3151  CA ASEP B1064      25.897   6.120  -0.774  0.50 18.88           C  
       ...

 (3) This renumbering also needs to be done for the LINK records:

       LINK         C   GLY B  63                 N  ASEP B1064     1555   1555  1.33  
       LINK         C  ASEP B1064                 N   VAL B  65     1555   1555  1.33  

Now BUSTER will understand that those two residues are different (B64
versus B1064) and that B1064 is linked from B63 and to B65. You can
check this by running

  MakeLINK -p test.pdb -o test.seq

and observing in test.seq

  ...
  RESIDUE B|62  ILE 63 XGPEPTIDE
  RESIDUE B|63  GLY 1064 GXPEPTIDE 64 GXPEPTIDE
  RESIDUE B|64  SER 65 PEPTIDE
  ...
  RESIDUE B|1064  SEP 65 PEPTIDE
  ...

One small issue still with that particular example (might not be
needed for yours): you need to add

  AnalyseGellySanityCheckForDuplicateBonds=no

to your command because there are now two peptide linkages from B63
(into B64 and B1064) resulting into some duplicated peptide-bond
restraints. This is something we will need to look at within BUSTER at
some point.

You don't need to run MakeLINK (this is just for testing) and should
be able to run 'refine' (BUSTER) directly in the normal way - e.g.

  refine -p test.pdb \
         -m 5za2.mtz \
         -autoncs -sim_swap_equiv \
         -l $CLIBD/monomers/n/NXL.cif \
         AnalyseGellySanityCheckForDuplicateBonds=no \
         -d 03 | tee 03.lis

> I could probably generate two chains of the same protein, assign 0.5
> occupancy to each, perhaps constrain all other residues by strict NCS and
> then upon refinement merge the chains back to generate proper file for
> deposition.  But this hack isn't just ugly, it's grotesque.
> 
> Does anyone know of any more elegant way to do this that ?

Not sure the above is more elegant, but it should work - maybe with
some adjustment for your particular case. And before deposition you
can then just move the B1064 residue up (to be beside the B64 one) and
edit the file back from B1064 to B64).

Cheers

Clemens & Andrew (for BUSTER developers)