Any instruction for conversion from relion to frealign?

Forums

Frealign

Hi, I consider trying frealign to continue to refine my 3d reconstuction after one run of relion 3d refine, I downloaded the conversion scripts and convert the data.star from last iteration to frealign par file, Now I am confusing what is the next step? which dataset I should use? where I put reference map, and which one ? what is the data format? star file or mrc file?
Please help me .
Thanks a lot,

Scripts

Axel Wed, 2015-05-06 14:43

Hi,

You can try the scripts on this page:
http://grigoriefflab.janelia.org/frealign_conversion_scripts

Note that the reliontofrealign script converts to a frealign 8 format (last time I checked anyway). I am not sure if there are scripts on the web site to handle conversion from Frealign 8 to Frealign 9, or if this is handled internally when using the utility submission and refinement scripts provided with the latest version of Frealign 9.

HTH,

Axel

I already downloaded it and

smfang Wed, 2015-05-06 14:48

I already downloaded it and now I need an example show me how I can use the dataset from relion for frealign. This is my first time to try to use frealign for 3d refinement. You know relion data is star file, Do you think I need to convert it to mrc stack file and also where I put reference map to?
Thanks a lot,
SHunming

Hello Shunming, Have you

Alexis Wed, 2015-05-06 14:52

Hello Shunming,

Have you already looked at the Frealign help pages? For example, this might be helpful:
http://grigoriefflab.janelia.org/frealign_getting_started

Alexis

Thanks a lot, Alexis, Now It

smfang Wed, 2015-05-06 15:01

Thanks a lot, Alexis, Now It is much clear for me. I will integrate all mrc stack files into one mrc stack. and then Try it according to website.
Shunming

Scripts

Axel Wed, 2015-05-06 14:54

Hi,

You can try the scripts on this page:
http://grigoriefflab.janelia.org/frealign_conversion_scripts

I just read your post in more detail:

1 - Frealign requires 3 inputs:

A stack of 2D images (the same one as you used in relion should work). If you are using the latest relion scripts, which generate a stack for each micrograph, you will have to generate a single stack with the particles maintained in the same order as in the parameter file.

A parameter file (generated by the reliontofrealign script)

A 3D reference map (which will be automatically generated given a particle image stack, and a parameter file)

2 - You should use the same dataset as you used in relion.

3 - The reference map is in mrc format, always (relion also generates mrc reference maps).

HTH,

Axel

PS Star files are parameter files, not image files. Images in Relion are stored as MRC files, as they are in Frealign. The reliontofrealign script converts the relion parameter file (a star file) to a frealign parameter file (a par file).
I would suggest you read through the Frealign and Relion papers and documentation to get a better idea where your data is, and how it relates to the reconstructions being generated. You may find that you are currently a bit confused about the process currently.

converting stack

sconnell Thu, 2015-05-07 01:13

Hi sorry for asking a follow up question but I though this might be a good place to ask some advice on making the image stack for frealign from a relion project. Basically I was just wondering if the following proceedure makes sence:

# generate a combined stack file frrom the star file
relion_stack_create --i STAR_FILE --o ROOT_STACKOUT

# normalize stack for frealign and flip polarity
e2proc2d.py STACKIN.mrcs STACKOUT.mrc --process=normalize.edgemean --mult=-1

Thanks in advance for any input.

I have not used these

niko Thu, 2015-05-07 12:51

I have not used these commands but this looks like it could work. Why not try it?

Fantastic!

Axel Thu, 2015-05-07 13:08

Hi,

I have an equivalent script, which works much slower than the one you wrote above.

Two comments though:

1 - The new Frealign does not require black particles on white background (now an option in the parameter file) and Relion normalizes the particles during extraction, so you can probably skip the e2proc2d portion (since working with stacks of 2D mrc images in eman2 usually requires the use of threedtothreed or threedtotwod or some such options)

2 - You can extract a single class from a multiple class star file (eg run....data.star) by using the following command, where the class assignment is value # 21:

awk '{if (NF<= 2) {print} else {if ($21==1) print }}' < run1_ct6_it017_data.star > class1.star

HTH

converting stack

sconnell Thu, 2015-05-07 14:56

Thanks for the confirmation. I have tried these out this last week and they are working. It is just the first time I worked with frealign in a few years and wanted to double check.

A few notes... I tested it with and without normalization and it appears to work both ways fine. I just noticed some deviation in the stddev of the images direct from relion and to make it "perfect" ran it through EMAN2. I inverted the polarity just to make it conform with the default value in mparameter. One problem with the process could be a large star files as relion_stack_create seems to work in memory

large files are problematic

Axel Thu, 2015-05-07 17:23

I agree that large stacks are problematic, I have been running into this issue myself.
The workaround is to either create a set of new stacks which have to be appended, or extract images individually.