Appion

Hi Neil,

I'm attaching the two aligned movies from which I showed the CTF confidence measurements.

16may27a_dyn_00027gr_00002sq_00003hl_00001en-a2.mrc - Original pixel size: 0.6A/pix
16may27a_dyn_00027gr_00002sq_00003hl_00001en-a.mrc - Binned by 2 after movie alignment: 1.2 A/pix

We are running myami/3.2:

URL: http://emg.nysbc.org/svn/myami/branches/myami-3.2
Repository Root: http://emg.nysbc.org/svn/myami
Repository UUID: 7194d2a7-0b6d-4db4-844f-13006c256e48
Revision: 19461
Node Kind: directory
Schedule: normal
Last Changed Author: acheng
Last Changed Rev: 19459
Last Changed Date: 2016-01-07 09:03:18 -0800 (Thu, 07 Jan 2016)

And, what we saw was that the CTFFIND3 confidence was lower on the binned by 2 image compared to the unbinned image. This is in these attached confidence images:

ctffind3_bin1 - Original pixel size: 0.6A/pix
ctffind3_bin2 - Binned by 2 after movie alignment: 1.2 A/pix

Microscope parameters:

• Original pixel size (super resolution mode): 0.6A/pix
• Cs = 2.7 mm
• kev = 200
• Magnification = 36000

Processing info:

• Data were collected as K2 movies on Talos-Artica microscope
• Movies are aligned using Appion
• For the binned micrograph, we specified a binning of 2 within the Create Frames page within Appion.

I suspect that this issue should be repeatable for anyone who bins their data using the motion correction software within Appion (along with myami version 3.2).

Do other Appion users always use the full-res movie for processing within Appion?

Let me know if I can help at all with this.

Thanks!
Mike

Based on the images attached, CTFFIND3 found different defocus values for each micrograph. 3.104um/3.138um in bin by 2 and 3.124um/3.133um in bin by 1. The Thon rings are the same, so I would expect a different result given a different defocus value.

Perhaps running it again will helps ctffind3 get the same value both times. As I see it it is a problem with CTFFIND3.

I could put in a hack that would allow us to measure the resolution for a given set of defocus value on both images to see if appion is consistent, but I would not expect appion to be consistent when CTFFIND3 provides different results.

Anchi any comments?

For Michael's question, those who use super-resolution tend to use bin=2 during alignment as it is done in Fourier space and should give better representation of the original. In most cases when information of the structure does not go close to physical Nyquist, people do not reprocess with bin=1 even after particle stack were cleaned up and ran through reconstruction. Those with only lower resolution information just use counting mode as there is no danger of aliasing.

I agree with Neil that if the CTF estimation program consider the two images different, there is nothing to fix on his side. As to why they gave different results, I can only speculate. Since all CTF estimation programs are done with dividing the image into square fields, the odd dimension of K2 images may cause it to include different fields in its processing and created different statistics in the diffractogram.
In addition, the extent of the Fourier space information that contributes to the background and normalization would be different in the two images, with bin=1 give more data point ( and noise) to fit.

Michael, please comment if you disagree.