Maybe something like this - where by default it just does a rigid body, but if
morph is set to “Y” it does a more extensive morph at three different radii:
set_refinement_immediate_replacement(1) #Accept rigid body without prompting.
mol_id=13 #Which molecule should we modify?
r1=20
r2=10
r3=5
r1_cycles=10
r2_cycles=5
r3_cycles=5
morph="N" #Do extensive morph?
do_rigid="Y"
if do_rigid=="Y":
for ch_id in chain_ids(mol_id): #Rigid body refine each chain first.
rigid_body_refine_by_atom_selection(mol_id, "//%s//"%(ch_id))
accept_regularizement()
if morph=="Y": #Sequential morph_chain at three radii.
for ch_id in chain_ids(mol_id):
for n1 in range(1,r1_cycles):
morph_fit_chain(mol_id, "%s"%(ch_id), r1)
for n2 in range(1,r2_cycles):
morph_fit_chain(mol_id, "%s"%(ch_id), r2)
for n3 in range(1,r3_cycles):
morph_fit_chain(mol_id, "%s"%(ch_id), r3)
set_refinement_immediate_replacement(0) #Return to default behavior.
On Dec 29, 2013, at 9:51 AM, Oliver Clarke <[email protected]> wrote:
> Thanks Paul - Initially I tried rigid body + 10 cycles of morph_fit_chain at
> a radius of 11 Å - that seemed to work better than rigid body alone, although
> some of the larger domain shifts are still not corrected.
>
> Perhaps I need a larger radius for that - I’ll try doing 10 cycles at a
> radius of 20 Å, followed by 5 each at 10 Å and 5 Å.
>
> Regarding blurred maps, is there any way to adjust the sharpening factor on
> the fly in a script, in such a manner that it can be returned to normal
> afterwards?
>
> Also, Coot seems to write out a copy of the pdb to the coot-backup directory
> after every cycle of morph_chain - is there any way to turn off this
> behavior? Normally it would be fine, but with a few hundred operations that
> adds up to several GB of backup files, which probably slows down the whole
> process somewhat.
>
> Cheers,
> Oliver.
> On Dec 29, 2013, at 6:21 AM, Paul Emsley <[email protected]> wrote:
>
>> On 29/12/13 02:51, Oliver Clarke wrote:
>>>
>>>
>>> I agree that normally rigid body refinement of each chain wouldn’t be very
>>> useful - I’m using this script in a situation where we have a large complex
>>> of many subunits that we need to fit to a quite low-resolution map (of a
>>> related complex from another species),
>>
>> Sounds familiar :-) (... as may be apparent at the Study Weekend).
>>
>>> and an initial round of rigid body fitting seemed to be helpful on
>>> individual chains but was quite laborious when performed manually for each
>>> chain.
>>
>> Eeek. Yes...
>>
>> Also consider Jiggle Fit.
>>
>>> This script seemed to help with that, and I was adding morph_fit_chain in
>>> the hope that it would correct errors in the curvature of helices, or the
>>> orientation of two domains within a multi-domain protein.
>>
>> OK, for the orientation of domains you need a larger radius, performed on
>> the chain. For the curvature of helices you need (I would imagine) a
>> smaller radius performed on a residue selection.
>>
>>>
>>> Yes, that was the problem - It works when I replace chain_id with ch_id.
>>>
>>
>> good.
>>
>>> So you would suggest calling morph_chain several times in succession?
>>
>> Definitely. That and the residue selection version.
>>
>>> That makes sense, because the modifications from one round were quite
>>> subtle. I’ll give it a go with 5-10 rounds and see how it works.
>>>
>>>
>>
>> Also, I found that fitting into blurred maps at the early stages (large
>> radius) was beneficial.
>>
>> It is not yet clear to me what's the best relationship of the round number,
>> resolution, and the radii.
>>
>> Paul.
>>
>>
>> <morph-residues-gui.scm>
>
