[phenixbb] rotamers

[Frank von Delft]

(oops, got carried away:  meant of course "of building MISSING 
SIDE-CHAINS AT OCC=0".  Embarrassing :)


On 24/03/2011 05:54, Frank von Delft wrote:
> Oh no!!  Please please please don't bring that hoary old canard of
> building in side-chains - it's Very Bad Practice, we don't build missing
> amino acids or ligands, do we, so why are sidechains different?  The
> "most likely rotamer" is still the WRONG rotamer:  if it were right,
> you'd see it.
>
> CNS used to do this, but as far as I could tell it was an artefact
> because CNS could not refine a residue if not all atoms were present, so
> to truncate something you had to call it ALA.  That's obviously wrong,
> and phenix fixed this, thank god, and good riddance.
>
> Now it's to come back...?  *sob*
>
> As for theoption (b):  that sounds like a very valid criterion for
> placing a side-chain.  In fact, it's awesome - so why set occupancies to
> zero?  Phenix already deals with validly modelled atoms very
> effectively:  set occ=1 and relax the B-factor restraints.
>
> (Sorry, you can tell I feel strongly... :)
>
> phx.
>
>
>
>
>
>
> On 23/03/2011 21:39, Pavel Afonine wrote:
>>     Hi Ed,
>>
>> you are asking the right question.
>>
>> We discussed the issue of residue side chains partially or fully lacking
>> density at PHENIX developers meeting last week. This was brought up by
>> an industrial consortium user.
>>
>> Some people tend to chop them off (which I do not not not like), some
>> let the refinement programs do what they want, and the programs stupidly
>> tends to stick these sidechains in whatever density is available around
>> (this is bad too).
>>
>> I think we decided that I have to do something about it, probably the
>> following:
>>
>> - if a side chain lacks the density and the full rotamer search does not
>> find any density, then a) set occupancies of affected atoms to zero, and
>> b) move the side chain to the highest probable rotameric state that
>> generates least clashes with surrounding atoms and may be still have
>> some good density fit (even though it is a bad density). I will work on
>> this. That's for sure.
>>
>> Pavel.
>>
>>
>>> Does refine restrains side chain rotamers to, say, the closest one from
>>> the rotamer dictionary or, perhaps, the starting coordinates (I find
>>> this less likely, of course)?  I am particularly interested in what
>>> would happen if a side chain is disordered and there is no strong
>>> electron density to support a specific conformation.  Do I understand
>>> correctly that the real-space refinement for the side chains is on by
>>> default, which means that disordered side chains would tend to stay
>>> within their biased density?
>>>
>>> I apologize if I missed the description of this in the manual, just
>>> point me to the right section then.
>>>
>>> Cheers,
>>>
>>> Ed.
>>>
>> _______________________________________________
>> phenixbb mailing list
>> phenixbb at phenix-online.org
>> http://phenix-online.org/mailman/listinfo/phenixbb
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb