Using unblur for individual particle alignment

Submitted by Matt I on Mon, 2015-10-19 13:51

Forums

Unblur & Summovie

Hi Tim et al,

I've been trying to use unblur on individual particles with a box size of 128x128 px. It appears to be performing poorly, the aligned particles look worse than a sum of the original unaligned particle stack. Is there a lower limit on the micrograph size for unblur to operate well?

Attached are sums of the original unaligned stack and the aligned stack for a particle.

Cheers,
Matt I

timgrant Mon, 2015-10-19 13:57

Hi Matt,

There must be enough signal in the raw frames to perform the alignment and I imagine that there simply isn't enough signal in your images for the alignment to work, hence you end up with more or less random shifts and bad sums. You would need (probably much) larger boxes to make this work. We typically only do whole frame alignments. In the DLP case I did align individual particles, but in this case the boxes were something like 1000A^2, and contained ~70 MDa of protein and RNA, so there was a good amount of signal.

Thanks,

Tim

rogelio Thu, 2015-10-22 20:30

Hi Tim,

I have few questions related to this post, I am planning to go back to my super res movies, correct for the magnification distortion, bin my movies and extract my particles using box files I already have. I want then to use unblur and summovie to correct the motion per particle. My questions are:

1.what program do you use to extract all frames in a movie given a box file? I have very big boxes of Microtubules (720 A) and want to give it a try to aligning individual particles as in the case of the DLP.
2. Also, what program do you use to fourier bin a movie? All my box files are relative to a binned by 2 sum images that I obtained with motioncorr, I would like to try using unblur for alignment but this does not have a binning option.

Does this strategy sound like a good option? Or should I just try the aligment and sum using the entire frame and then bin the sum image to extract again my particles?

Thanks a lot!

timgrant Fri, 2015-10-23 11:12

Hi,

The latest frealign comes with a program for Fourier binning written by Alexis Rohou. It is called resample.exe or resample_mp.exe for the threaded version (not rsample.exe!). You should be able to use that to do all your Fourier binning. The latest versions of unblur outputs the shifts in angstroms, so it is easy to align with one pixel size and sum with another.

For doing my particle alignment, I used a hacked version of unblur which took movies and co-ordinates as input and output aligned particle sums. This is quite old now though, and doesn't have the latest updates - plus it would not work directly with boxer co-ordinates, so you may be better extracting your particles another way (e.g. with EMAN2, or IMAGIC) and aligning them directly.

As another heads up, if you are doing the magnification distortion on your movies, then your old co-ordinates will no longer be in the correct place, the correction will move things by quite a lot over the scale of a whole image.

Cheers!

Tim