Hi Bettina,
The out-of-plane sampling rate is determined by the --healpix_order
argument. I can't see this in your command line, so assume it's the
default of 2 (15 degrees). That would explain your results. Use 4 or 5
instead. If you use the GUI to generate your arguments, this should not
happen.
HTH,
Sjors
> Hi Sjors,
>
> Thank for your respond.
>
> I am averaging microtubule doublets and triplets. After auto refinement
> and
> a round of 3D classification I realized that some of my subvolumes, about
> 30%
> are upside down. So I flipped them by redefining the angles in the
> data.star file.
> Then I flipped the data.star file into originally well aligned and upside
> down parts,
> and calculated the average for both using 3D classification with one class
> and
> --skip_align option to make it sure it worked. The two averages are close
> but
> slightly rotated.
> So I tried to align the second half to the average of the first using 3D
> classification,
> one class and limited angular range one iteration only. The orientation of
> the second set's average got closer to the first, but it looks worse.
> I checked the data.star files and the angles changed more then the
> specified range in
> one iteration only and looks noisier. I enclosed two figures rot vs
> subvolume number to show this. I would not worry if the quality of the
> average did not change.
>
> It is possible that I do not know how to restrict the angles correctly, so
> here is my
> command line and the out put of the run
>
> `which relion_refine` --o Class3D_new3/run3 --i
> Class3D_new3/run2_it001_data.star --particle_diameter 800 --angpix 8.16
> --ref ClassI_new.mrc --ini_high 30 --no_parallel_disc_io --ctf
> --ctf_corrected_ref --iter 1 --tau2_fudge 4 --K 1 --flatten_solvent
> --zero_mask --strict_highres_exp 30 --oversampling 1 --psi_step 1
> --sigma_ang 1.66667 --offset_range 2 --offset_step 1 --sym C1 --norm
> --scale --j 1 --memory_per_thread 4
> Running in double precision.
> Estimating initial noise spectra
> 1.77/1.77 min
> ............................................................~~(,_,">
> Estimating accuracies in the orientational assignment ...
> 1/ 1 sec
> ............................................................~~(,_,">
> Auto-refine: Estimated accuracy angles= 0.33 degrees; offsets= 0.195
> pixels
> CurrentResolution= 30.2222 Angstroms, which requires orientationSampling
> of at least 4.28571 degrees for a particle of diameter 800 Angstroms
> Oversampling= 0 NrHiddenVariableSamplingPoints= 363
> OrientationalSampling= 15 NrOrientations= 11
> TranslationalSampling= 1 NrTranslations= 33
> =============================
> Oversampling= 1 NrHiddenVariableSamplingPoints= 23232
> OrientationalSampling= 7.5 NrOrientations= 88
> TranslationalSampling= 0.5 NrTranslations= 264
> =============================
> Estimated memory for expectation step > 0.964804 Gb, available memory =
> 4
> Gb.
> Estimated memory for maximization step > 0.116198 Gb, available memory =
> 4
> Gb.
> Expectation iteration 1 of 1
> 18.25/18.25 min
> ............................................................~~(,_,">
> Maximization ...
> 15/ 15 sec
> ............................................................~~(,_,">
>
> I am not sure how to interpret the output_sampling.star file, but it
> does not look like it limited the angular range to +/- 5 degree in 1
> degree
> increments. Here it is:
> data_sampling_general
>
> _rlnIs3DSampling 1
> _rlnIs3DTranslationalSampling 1
> _rlnHealpixOrder 2
> _rlnSymmetryGroup C1
> _rlnTiltAngleLimit -91.000000
> _rlnPsiStep 1.000000
> _rlnOffsetRange 2.000000
> _rlnOffsetStep 1.000000
> _rlnSamplingPerturbInstance -0.212801
> _rlnSamplingPerturbFactor 0.500000
>
>
> data_sampling_directions
>
> loop_
> _rlnAngleRot #1
> _rlnAngleTilt #2
> 45.000000 80.405932
> 56.250000 70.528779
> 33.750000 70.528779
> 45.000000 60.000000
> 67.500000 60.000000
> 78.750000 48.189685
> 56.250000 48.189685
> 75.000000 35.659088
> 22.500000 60.000000
> 33.750000 48.189685
> 11.250000 48.189685
> 15.000000 35.659088
> 45.000000 35.659088
> 67.500000 23.556464
> 22.500000 23.556464
> 45.000000 11.715852
> 135.000000 80.405932
> 146.250000 70.528779
> 123.750000 70.528779
> 135.000000 60.000000
> 157.500000 60.000000
> 168.750000 48.189685
> 146.250000 48.189685
> 165.000000 35.659088
> 112.500000 60.000000
> 123.750000 48.189685
> 101.250000 48.189685
> 105.000000 35.659088
> 135.000000 35.659088
>
>
> Thanks for you help,
>
> Bettina
>
>
>
>
>
> On Wed, Oct 5, 2016 at 12:16 AM, Sjors Scheres <[log in to unmask]>
> wrote:
>
>> Hi Bettine,
>> Local searches are centred around the best rotations and translations
>> from
>> the previous iteration. Therefore, in multiple iterations, particles can
>> move more than you specified.
>> HTH,
>> Sjors
>>
>> > Hi all,
>> >
>> > I am doing subtomogram averaging using relion14. I am trying to
>> > push the resolution using Class3D with one class with limited search
>> > range.
>> > Unfortunately both the translations and the angles are changing more
>> than
>> > the allowed range and the parameters are drifting.
>> >
>> > Does anyone had the same experience?
>> >
>> > Any suggestion is appreciated. Thanks,
>> >
>> > Bettina
>> >
>>
>>
>> --
>> 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
>>
>>
>
--
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
|