Hi David,
Great, thanks! That's very useful information.
Regards,
-jh-
On 2/8/19 6:25 PM, David Boyer wrote:
> I think you are correct about the last point. Although, I don't know
> if it is necessarily "overfitting" just because you are manually
> controlling tau...(it may *lead* to overfitting).
>
> I usually start to refine twist when 3D classification hits 6-7
> Angstrom, then let it stabilize (rise may vary too much if refined
> before strands show up in z-direction). Then push to ~4-4.8 Angstrom
> with manual control of healpix/offset/tau to refine the rise. Once
> twist and rise are stable (and you've pruned the data through multiple
> Class2D and Class3D jobs), people generally end with AutoRefine as
> published... In my experience, twist and rise do not vary *too* much
> from initial guesses (refined pitch between 0-100 A from initial guess).
>
> Also, to overcome not having enough memory for K=4,5,6... can run
> iterative K=3 jobs, taking the best class(es) from each run and having
> them as input to next K=3 job.
>
> Running with K=3 and a search for helicity can also help separate
> particles that have the same underlying fold, but have a different
> helicity. I had one case where the pitch of one fibril was 900A and
> another fibril was 1000A (the x-y slice looked nearly identical). I
> was able to separate them out nicely in Class3D and am now testing if
> there is any effect on resolution for a reconstruction coming from a
> mixture of these particles, versus reconstructions from each
> subset...(Fitzpatrick et al, supposes not and use z_percentage 0.1 to
> minimize any effect a mixture of fibrils with slightly different
> pitches might have.) Just something else to be aware of I guess.
>
> On Fri, Feb 8, 2019 at 4:42 PM Heumann John <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Hi David,
>
> Thanks for your suggestions.
>
> My runstring arguments were very similar to yours, and it looks
> like the
> limiting factor in my case is indeed the number of classes. On my
> system
> (not a cluster, as you've guessed) 1-3 classes work, but 4 or more do
> not. (Previously, I'd been trying 5 - 20).
>
> That's an interesting observation about --bimodal_psi not being in
> the
> default runstring for helical reconstruction. Looking at the
> refine_mpi
> source code, it looks to me like this may not be necessary (and
> may in
> fact be ignored) when "Apply helical symmetry" is on. Regardless of
> whether I add this as an Additional argument, I see output like
> > Number of helical segments with psi angles similar/opposite to
> their
> > priors: 3741 / 3465 (48.0849%)
> suggesting that bimodal priors are in use. Perhaps one of the Relion
> folks can confirm this.
>
> Re my 2nd question, it wasn't so much the use of particles from
> multiple
> 2D classes which I was confused about. What I was wondering was why
> choose to run 3D classification with a single output class rather
> than
> 3D auto-refinement? The only rationale I've come up with is that 3D
> classification allows you to manually control the amount of
> regularization... in this case intentionally overfitting to try to
> bring
> out features in z to help chose pitch and rise. Is that really why
> this
> was done this way, or is there some other reason I've overlooked?
>
> Thanks again for your help!
>
> Regards,
> -jh-
>
> --
> John M. Heumann
> Department of Molecular, Cellular, and Developmental Biology
> 347 UCB, University of Colorado
> Boulder, CO 80309-0347
>
> On 2/8/19 1:37 PM, David Boyer wrote:
> > Hi John,
> >
> > In terms of memory issues, you should be fine using more than 2
> > classes (I usually use 3). Below is a sample command line from a
> K=3
> > job for amyloid fibrils (don't forget to use --bimodal_psi, last
> time
> > I checked it wasn't in the Relion GUI for Class3D on the Helix
> tab...)
> >
> > /home/davboyer/cryoem/openmpi-3.0.0/build/bin/mpiexec --bind-to
> none
> > `which relion_refine_mpi` --o Class3D/rod_320_K3_R2/run --i
> > ./Select/rod_refine_class2_particles/particles.star --ref
> > Class3D/refine_rod_K1/run_179p355_K1_ct113_it125_class001.mrc
> > --ini_high 40 --dont_combine_weights_via_disc --scratch_dir
> /scratch
> > --pool 30 --ctf --ctf_corrected_ref --iter 25 --tau2_fudge 4
> > --particle_diameter 336 --K 3 --flatten_solvent --oversampling 1
> > --healpix_order 3 --offset_range 5 --offset_step 2 --sym C1 --norm
> > --scale --helix --helical_outer_diameter 200 --helical_nr_asu 14
> > --helical_twist_initial 179.352996 --helical_rise_initial 2.407172
> > --helical_z_percentage 0.3 --sigma_psi 5 --j 5 --gpu ""
> --bimodal_psi
> > --limit_tilt 30
> >
> > In this case I was using SLURM for working on our cluster, but to
> > adapt for a single machine (perhaps what you have according to your
> > description?) we could say that I was using 3 mpi tasks per node
> (16
> > cores and 2 1080's on each node), so there is one master and two
> > slaves on the node. I also gave each mpi slave 5 cpus (hence j 5)
> > therefore each gpu card talks to 5 cpus (could do more at the early
> > stage of classification when sampling is less memory intensive). If
> > using a single machine, I would suggest something like mpiexec -n 3
> > --bind-to-none .... j 5. If you move towards higher healpix, you
> may
> > need to turn down the j number to 4, 3, 2, or 1 so the gpus
> don't run
> > out of memory.
> >
> > Second question, yes *particles* from multiple 2D classes are
> put into
> > 3D classification to supply all the views of the helix. So you
> just do
> > a selection job from your 2D classes to gather all the particles
> that
> > contribute to the same species and are in higher resolution
> classes as
> > input for your 3D job. Unless you are working with big boxes
> where all
> > the views of the helix are present, this is necessary for IHRSR.
> >
> > Good luck!
> >
> > David
> >
> >
> > On Thu, Feb 7, 2019 at 4:19 PM John Heumann
> <[log in to unmask] <mailto:[log in to unmask]>
> > <mailto:[log in to unmask]
> <mailto:[log in to unmask]>>> wrote:
> >
> > I'd appreciate some guidance regarding helical refinement?
> > Specifically:
> >
> > 1) I'm trying to refine some amyloid data using parameters like
> > those used by Fitzpatrick et al for Tau (box size 280, ~10%
> > spacing), but seem to be continually running into segmentation
> > violations or GPU memory allocation errors particularly
> during 3D
> > classification. My presumption is that the segmentation
> violations
> > also result from an out-of-memory issue, but on the main
> computer
> > instead of the gpu. This is on a system with 128 GB of ram and 4
> > GTX 1080's. So far, the only thing that seems to help is
> reducing
> > the number of classes to 2, but that largely defeats the purpose
> > of classification. Reducing the number of mpi processes and / or
> > threads seems ineffective. Can some please describe the main
> > determinants of memory usage during helical refinement? I assume
> > small box sizes might help, but that would also lead to reduced
> > overlap and more particles.
> >
> > 2) I'm having a hard time understanding the following
> portion of
> > the Fitzpatrick et al Methods:
> >
> > "We then selected those segments from the PHF and SF
> datasets that
> > were assigned to 2D class averages with β-strand separation for
> > subsequent 3D clas-
> > sification runs. For these calculations, we used a single
> class (K
> > = 1); a T value of 20; and the previously obtained
> sub-nanometre
> > PHF and SF reconstructions,
> > lowpass filtered to 15 Å, as initial models"
> >
> > So wait, multiple 2D classes are input to 3D classification with
> > only a single output class? What purpose does that serve?
> Was this
> > done solely to generate a 3D alignment for exploring twist /
> rise
> > parameters as is described next?
> >
> > Thanks!
> >
> > Regards,
> > -jh-
> >
> >
> ########################################################################
> >
> > To unsubscribe from the CCPEM list, click the following link:
> > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
> >
>
--
John M. Heumann
Department of Molecular, Cellular, and Developmental Biology
347 UCB, University of Colorado
Boulder, CO 80309-0347
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|