Dear Sjors,
I followed your described workflow. I selected the respective class
from the masked classification of the subtracted particles, reverted
back to the non-substracted ones and did the refine-3D with 1.8
degrees angular sampling. The result looks in principle ok, but the
reported resolution (7.7 A) is worse than the one reported in the
Class3D job (5.8 A) and during post-processing I get a warning, that
the unmasked FSC extends beyond the solvent-corrected FSC, the
reported resolution gets even worse (8.1 A) and the FSC plot looks
pretty strange (the FSC corrected curve drops below -0.5 before it
converges to 0).
I would be very thankful for some suggestions.
Best,
Julian
Zitat von Sjors Scheres <[log in to unmask]>:
> Dear Julian,
>
> More classes just gives it more space to separate out, as you do not align
> anyway, the increase in number of parameters is not large.
> The higher T fudges for the observation that the signal now lives in a
> small part of the box, while it is measured as its power from the whole
> box.
>
> You say refinement has failed. You could try to select your class from the
> expanded classification, and then perform a auto-refinement where you only
> perform local angular searches: that is, use a start sampling that is just
> as large as the one from where you do local searches, for example both at
> 1.8 degree. The input map for that refinement could be the class you had
> identified. Doing only local searches will prevent particles 'loosing
> their way again' if you start from a low-resolution initial model. This
> whole procedure also works with signal subtraction, but make sure to
> revert back to the non-subtracted particles before the 3D refinement (i.e.
> only use subtracted images in the classification).
>
> HTH,
> Sjors
>
>
>
>> Dear Sjors and all the others,
>>
>> thank you very much for the very helpful input!
>>
>> We now expanded the star file and made a masked classification which
>> resulted in one class nicely showing the expected conformational
>> change. However, refinement of this class has led to a bad result.
>>
>> We will now do the particle substraction as you suggested and try again.
>>
>> For my understanding, could you clarify the need of many classes and
>> the higher-than usual T-factor?
>>
>> Best,
>> Julian
>>
>>
>> Zitat von Sjors Scheres <[log in to unmask]>:
>>
>>> Dear Julian,
>>> It is as Bayly says:
>>>
>>> relion_particle_symmetry_expand --i refine3D/jobXXX/run_data.star
>>> --sym C3 --o expanded_particles.star
>>>
>>> Then there are multiple options. I would suggest making a mask
>>> around 1 monomer and another mask around the other 2 monomers. Then
>>> subtract 2 monomers from all expanded particles, and feed the
>>> resulting images into a 3D classification (without re-alignment)
>>> that uses the single-monomer mask. Use many classes (10?) and a
>>> higher-than usual T-factor, say 10-20.
>>>
>>> The original reference is :
>>> S.H.W. Scheres (2016)Â "Processing of Structurally Heterogeneous
>>> Cryo-EM Data in RELION" Meth. Enzym. 579, 125-157 (section 4.7)
>>>
>>> HTH,
>>> Sjors
>>>
>>>
>>> Charles Bayly-Jones wrote:
>>>> Yes, the relion_particle_symmetry_expand function will allow you to
>>>> accomplish this. Then perform masked 3D classification as you
>>>> described. It is very useful indeed :)
>>>>
>>>> You could also use the localised reconstruction method from Ilca et
>>>> al. (2015) doi:10.1038/ncomms9843
>>>> <http://www.nature.com/ncomms/2015/151104/ncomms9843/full/ncomms9843.html>
>>>> Best of luck,
>>>> Charlie
>>>> _________________*
>>>> *
>>>>
>>>> *Charles Bayly-Jones*
>>>> BSc(ScSchProg)(Hons)
>>>>
>>>> *PhD Candidate*
>>>> *Teaching Associate*
>>>> *
>>>> *
>>>> *Monash Biomedicine Discovery Institute*
>>>> *Department of Biochemistry and Molecular Biology*
>>>> Level 2, Building 77
>>>> 23 Innovation Walk
>>>> Clayton VIC 3800
>>>> Australia
>>>>
>>>>
>>>> On Tue, 5 Mar 2019 at 21:01, Julian Reitz <[log in to unmask]
>>>> <mailto:[log in to unmask]>> wrote:
>>>>
>>>> Dear all,
>>>>
>>>> we are working on a dataset from a homo-trimeric protein. Until
>>>> now we have very successfully refined the structure to good
>>>> resolution applying C3 symmetry in RELION.
>>>>
>>>> However, we also have a sample from this protein that contains a
>>>> low affinity binding substrate and we would like to do a
>>>> classification to see if there are some subunits showing the
>>>> expected conformational change due to substrate binding. The point
>>>> is, that we do not expect the ligand to bind to all three monomers
>>>> of one particle but that the conformational change might only
>>>> happen in one of the monomers of a particle. Therefore, normal
>>>> 3D-classification seems unrewarding and a masked classification
>>>> would suffer from the aspect, that we always only look on one of
>>>> the monomers ignoring two-thirds of the data.
>>>>
>>>> We were thinking that it might be helpful if we somehow triplicate
>>>> the particles, each rotated by 120° to get all of the three
>>>> monomers from one particle inside the mask. Is there any clever
>>>> way to do this in RELION or does anybody else has a good idea how
>>>> we could proceed?
>>>>
>>>> Best,
>>>> Julian
>>>>
>>>>
>>>> ########################################################################
>>>>
>>>> To unsubscribe from the CCPEM list, click the following link:
>>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>>>>
>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>> To unsubscribe from the CCPEM list, click the following link:
>>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>>>>
>>>
>>> --
>>> 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
>>>
>>> ########################################################################
>>>
>>> To unsubscribe from the CCPEM list, click the following link:
>>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>>
>>
>> --
>> Julian Reitz, Dipl.-Biochem.
>> J.W. Goethe Universität Frankfurt am Main
>> BMLS - Buchmann Institut for Molecular Life Sciences
>> Frangakis Group
>> Max-von-Laue-Str. 15 Raum 1.658, D-60438
>> Tel: +49 (0)69 798-46429
>> e-mail: [log in to unmask]
>> http://fcem.uni-frankfurt.de/em
>>
>> ########################################################################
>>
>> To unsubscribe from the CCPEM list, click the following link:
>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
>>
>
>
> --
> 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
--
Julian Reitz, Dipl.-Biochem.
J.W. Goethe Universität Frankfurt am Main
BMLS - Buchmann Institut for Molecular Life Sciences
Frangakis Group
Max-von-Laue-Str. 15 Raum 1.658, D-60438
Tel: +49 (0)69 798-46429
e-mail: [log in to unmask]
http://fcem.uni-frankfurt.de/em
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|