Hi Huy,
bin1 will indeed be faster than bin2...
You could provide a upscaled reference at higher resolution, and a
data.star file with the correctly rescaled rlnDtectorPixelSize, as well as
rlnOriginOffsetsX/Y. Details have been discussed on this list. If you do,
you can then start the refinement already from local angular searches.
That should speed up things.
Remember that rescaling does not need to be an integer factor. Any even
box-size will do, so you could also try for example a box of 400x400.
HTH,
Sjors
> Dear list,
>
> Anyone has any recommendation for routine going from bin 2 to bin 1
> dataset? My workstation has only 128GB RAM and my box size is 512 x 512
> unbinned. Using bin2 dataset (256 x 256), I already hit Nyquist frequency
> resolution. Therefore, I need to go to unbinned dataset. From Relion FAQ,
> it says that using bin1 will not increase the processing time. However,
> the
> refinement is going so slow and crash a lot of time due to lack of RAM.
>
> Is there anyway to continue the refinement of unbinned without going
> through all the iteration from 60A to high res? Can I refine from the last
> iteration of the bin 2 data further?
>
> I saw that it seems like I can modify the optimiser.star file as the
> starting point for bin1 data set. Is this recommended to do so?
>
> Best,
> Huy
>
--
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
|