Hi Joel,

On 01/08/2014 04:39 PM, J. Meyerson wrote:
> Hi Sjors,
> Great, that answers my question. I have three followup questions
> though. I hope this isn't too much of a barrage...
>
> First, I wish to do the masking at the 3D refinement stage, so which
> file should I extract the side views from? Perhaps the star file that
> contains particles from the 3D classes of interest?
>
I don't understand this: why would you want to extract side views?
You'll mask the 3D reference at every iteration: the 2D particles are
not masked with this 3D mask. If you want to do local angular searches
as I mentioned in my previous reply, then just input the last data.star
file from the unmasked refinement as the particles star file in the new
refinement.

> Second, can you help me understand which phi and theta values I should
> extract, such that only side views are considered in the alignment? It
> seems to me that I should extract all angles phi, and only theta
> values of 90 degrees +/- 10 degrees or so, but I suppose this all
> depends on how Relion orients the map.
For conventions see:
http://www2.mrc-lmb.cam.ac.uk/relion/index.php/Conventions_%26_File_formats#Euler_angle_definitions
It all depends on the orientation of your molecule w.r.t the XYZ axes
and what you call a side view.

>
> Third, how soft would you recommend I make the spherical mask? I plan
> to use Bsoft's "beditimg" utility, which has a Gaussian "edgewidth"
> option (also, as you indicate, I'll set the "fill" option as 1, and
> "background" as 0).
We often put 5 pixel soft edges on our masks. BTW: The next release of
relion will have a program to help you with this. But that doesn't help
you now, as we don't distribute it yet: keep your eyes open for a
beta-release on the relion wiki during some time this Spring.

HTH, S

>
> Thanks,
> Joel
>
> On 2014-01-08 08:29, Sjors Scheres wrote:
>> Hi Joel,
>> We have had good results doing exactly what you suggest: make a soft
>> mask
>> around the region that you want to improve (it may be any shape, but
>> values should be in [0,1], 1 being what you want to keep, 0 being
>> what you
>> want to mask away). You can provide this 3D mask on the auto-refine (or
>> Class3D) GUI.
>> If the masked out region is small compared to the entire molecule
>> then it
>> may be better to perform only local angular searches around angles
>> assigned without any mask.
>> HTH, S
>>
>>
>>
>>> Hi Sjors,
>>> I'm working with a membrane protein and the structure shows
>>> better-resolved features in the extracellular portion, compared to the
>>> transmembrane portion. I believe that the resolution of the
>>> transmembrane region is compromised because of a combination of two
>>> factors. First, most of the protein mass is extracellular and that
>>> region thereby dominates the alignment. Second, there might be some
>>> degree of flexibility where the extracellular portion of the protein
>>> meets the transmembrane portion.
>>>
>>> Can you suggest a strategy to improve the resolution in the
>>> transmembrane region? Perhaps using a soft-spherical mask around this
>>> region during 3D refinement? It's important to note that the protein
>>> essentially looks like a rod protruding from the membrane, and I have
>>> good angular coverage of the side views.
>>>
>>> Thanks,
>>> Joel
>>>

--
Sjors Scheres
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge Biomedical Campus
Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 267061
http://www2.mrc-lmb.cam.ac.uk/groups/scheres