To make this easier, you can use the particle_symmetry_expand utility in relion 2, I believe it has options for helical symmetry

Cheers
Oli

> On Dec 7, 2016, at 7:38 AM, Jesus GOMEZ <[log in to unmask]> wrote:
> 
> Hi, Julian,
> 
> One thing you could try is to perform the subtraction Sjors sugests with each of the 22 subunits separately, which will give you 22 separate datasets each featuring a single segment in you helix. If you merge all these into a single dataset you could perform 3D classification and separate the particles/subunits according to their structural heterogeneity, disregarding their original position in the helix/boxes. The principle is the same as imposing the symmetry, but you focus in a single segment. This should make the classification easier, since you avoid the problem you anticipate in your previous message, having different states randomly distributed along your helix. Probably recentering these subtracted particles would work best for the classification. Then you could also go back to the original coordinates and do things like using only the particles with one given conformation to reconstruct a full helix in that conformation (given that the conection between segments is conserved in the subtracted particles) and apply the helical symmetry to that one, since as Sjors says, "will work much better if all subunits in little stretches of filaments adopt the same conformation". That said, I have no experience using Relion for helical reconstructions, so take my advice with care and ideally have someone with more experience confirm that this is a reasonable idea.
> 
> Hope that was of some help, best regards,
> 
> Jesus
> 
> 
> 
> 
>> On 2016-12-07 12:42, Sjors Scheres wrote:
>> Hi Julian,
>> That's a much more difficult problem! You could try and make a mask
>> around a single subunit, and subtract the rest of the filament, in
>> order to then reconstruct different classes of that individual subunit
>> only. However, whether you then impose helical symmetry or not is a
>> difficult question. Probably, the classification will work much better
>> if all subunits in little stretches of filaments adopt the same
>> conformation, but that's probably not the case.
>> From our own experience, I can say we have had no success at all in
>> classifying independently floppy parts of helices... Our
>> classifications usually only distinguish good from bad segments, or
>> separate out filaments with slightly different helical twist and rise
>> (which can be refined independently for different classes).
>> HTH,
>> Sjors
>>> On 12/07/2016 11:35 AM, Julian Reitz wrote:
>>> Dear Sjors,
>>> thank you for your answer. Indeed the chapter in Meth Enzym was from great help.
>>> But I figured out a problem for our case that for me seems to be more severe.
>>> We are working on a helical reconstruction.
>>> Considering the relevant parameters we have approx. 22 subunits in our particle boxes.
>>> If I now mask out a single region in the reference map I only use the information of one of these 22 subunits per particle. And if I mask out the region in all the subunits I might have the problem that they are not all in the same state and that the structural heterogeneity is not consistent for all subunits in one particle.
>>> What would be your advise to deal with this?
>>> Thanks,
>>> Julian
>>> Zitat von Sjors Scheres <[log in to unmask]>:
>>>> Dear Julian,
>>>> Hopefully our most recent chapter in Meth Enzym will help. http://dx.doi.org/10.1016/bs.mie.2016.04.012
>>>> Indeed, it is the 'Reference mask' option that is needed, possibly together with the subtraction jobtype.
>>>> For making masks, you can use relion_mask_create, but we also often use other software (e.g. volume eraser in Chimera, a fitted PDB-file, etc)
>>>> HTH,
>>>> Sjors
>>>>> On 12/05/2016 01:10 PM, Julian Reitz wrote:
>>>>> Hi all,
>>>>> we would like to do some 3D-classifications of our data.
>>>>> However we have reasons to believe, that there are positions in the structure that show independent heterogeneity. Meaning, that it is difficult to cover this variety with a normal classification.
>>>>> We would like to do something as follows:
>>>>> We would like to apply a mask that masks out a single of this regions and that the classification (and the following refinement) puts the focus on this region.
>>>>> As far as there is no misunderstanding on my site this is not what the "Reference mask" option for 3D-Classification is made for, since it should be used to mask out protein against the solvent.
>>>>> Any suggestions about how to do this?
>>>>> Best,
>>>>> Julian
>>>> -- 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