[3dem] Large image box size in CryoSparc2 and Relion3

[Daniel Asarnow]

Elena,
To back up Takanori here, even in 3D, if you downsample to a pixel size of
say 3.5 Å, it would be like setting the E step limit in Relion or the
refinement limit in Frealign/cisTEM to o 7 Å. I have a number of datasets
where aligning only to 7 Å achieves 4 Å or better resolution (of course
pixel size has to be small enough to tell). You can take the angles learned
for the coarse images and then use them to reconstruct the original images
and see how high the resolution goes. There's no need for the very high
resolution information to be used in alignment unless you know the data
will go to *very* high resolution.

Best,
-da

On Mon, Oct 22, 2018 at 9:50 AM Takanori Nakane <tnakane at mrc-lmb.cam.ac.uk>
wrote:

> Hi,
>
> > I do not want to coarsen them.
>
> Really? For Class2D, you don't need high resolution at all.
> Down-sampling to 3.5 A / pix is the best way forward.
>
> For Refine3D, you should specify "Skip padding: Yes". With this option,
> we have successfully refined a 1024 px box using GTX 1080 Ti (11 GB / GPU).
>
> Best regards,
>
> Takanori Nakane
>
> > Hi Everybody,
> >
> > I wonder if you can help me, any suggestions are welcome:
> >
> > We need to process particles that are big enough and fit only in the
> > boxes of the size 900x900. I do not want to coarsen them.
> >
> > Is there any way I can process them in CryoSparc (ver 2), or Relion 3?
> > So far? CryoSparc2 crashes at 2D classification complaining that :
> > *cufftAllosFailed*
> >
> > If I want to have 100? or 20 classes it fails in ~5min, with 5 classes
> > it failed after iteration #19. The error message is the same.
> >
> > Relion 3 was more efficient: running two (2) month for 50 classes?
> > reaching the iteration #2. Then the guy was mercilessly killed.
> >
> > I am using 4 GPU with the memory of ~ 12Gb each
> >
> > Set consists of ~ 35 000 images.
> >
> > Many thanks in advance. Elena
> >
> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > Elena Orlova
> > Crystallography, Institute for Structural and Molecular Biology,
> > Department of Biological Sciences
> > Birkbeck,  University of London,
> > Malet Street,
> > LONDON  WC1E 7HX
> > Tel: (+44) (0)20 7631 6845
> > Fax: (+44) (0)20 7631 6803
> > e-mail: e.orlova at mail.cryst.bbk.ac.uk
> >
> > _______________________________________________
> > 3dem mailing list
> > 3dem at ncmir.ucsd.edu
> > https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
> >
>
>
> _______________________________________________
> 3dem mailing list
> 3dem at ncmir.ucsd.edu
> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20181022/3c28ec90/attachment-0001.html>