Project

General

Profile

Actions

Bug #3550

closed

CTFFIND failing for some micrographs

Added by Emily Greene almost 9 years ago. Updated over 8 years ago.

Status:
Closed
Priority:
Normal
Category:
-
Target version:
-
Start date:
09/12/2015
Due date:
% Done:

0%

Estimated time:
1.00 h
Spent time:
Affected Version:
Appion/Leginon 3.2
Show in known bugs:
No
Workaround:

Change field size from 512 pixels to 256 pixels


Description

CTFFIND and Ctftilt are failing for seemingly random micrographs in datasets with the following error:

Starting image 1 ( skip:54, remain:776 ) id:3259216, file: 15sep10a_01sq
... Pixel size: 723.276
Found 0 ctf values
!!! WARNING: no best CTF value for image 15sep10a_01sq
... Xmag=69, dstep=5.00e-06, mpix=7.23e-08
Defocus search range: 28738951 A to 28788951 A (2873.90 to 2878.90 um)
... running ctf estimation at Sat Sep 12 10:32:09 2015
/data2/log-l/progs/myami/appion/bin/ctftilt64.exe
15sep10a_01sq.mrc
15sep10a_01sq-pow.mrc
2.7,300.0,0.07,69.129903384,5.0,2
512,100.0,10.0,28738951.6,28788951.6,500.0,500.0,-0.00249969375407,10
... ctf estimation completed in 429.47 msec
Traceback (most recent call last):
File "/data2/log-l/progs/myami/appion/bin/ctfestimate.py", line 449, in <module>
imgLoop.run()
File "/data2/log-l/progs/myami/appion/appionlib/appionLoop2.py", line 86, in run
results = self.loopProcessImage(imgdata)
File "/data2/log-l/progs/myami/appion/appionlib/appionLoop2.py", line 136, in loopProcessImage
return self.processImage(imgdata)
File "/data2/log-l/progs/myami/appion/bin/ctfestimate.py", line 317, in processImage
( nominaldf, self.ctfvalues['defocusinit']))
KeyError: 'defocusinit'

It then crashes and the micrograph needs to be hidden to process any other images.

Example command: ctfestimate.py --runname=ctffindrun1 --rundir=/data2/log-l/appion/15sep10a/ctf/ctffindrun1 --commit --projectid=6 --session=15sep10a --no-rejects --no-wait --continue --ampcarbon=0.15 --ampice=0.07 --fieldsize=512 --medium=ice --bin=2 --resmin=100 --resmax=10 --defstep=500 --numstep=50 --dast=500 --ctftilt --expid=14176 --jobtype=ctfestimate


Related issues 1 (0 open1 closed)

Has duplicate Appion - Bug #4437: ctftilt not workingDuplicateZhening Zhang09/10/2016

Actions
Actions #1

Updated by Emily Greene almost 9 years ago

  • Subject changed from CTF failing for some micrographs to CTFFIND failing for some micrographs
Actions #2

Updated by Dmitry Lyumkis almost 9 years ago

Emily, I had a look at this issue:

It looks like it’s not an Appion issue, and it's not the latest CTFtilt version, but rather the micrographs themselves. The latest version of CTFTilt seg-faults on this micrograph as well. I did a prime number decomposition for the image size (3710x3838), which provides ([2,5,7,53] and [2,19,101]). This isn’t ideal for FFTs. I clipped the micrograph to 3072x3072 (prime numbers [2,3] only, 2**10*3). This works fine. So the bug seems to be inherent to CTFTilt. We should come up with some way to deal with this in Appion … one way is to preprocess the images & clip them. This may not be ideal, because we will have to account for the particle coordinates differently. I suppose this might be a temporary solution. YZ, you might have a better idea of how to deal with the problem, as you’ve been applying the formulae to your picks to obtain the actual CTFs. Thoughts?

Actions #3

Updated by Neil Voss almost 9 years ago

Is this for both CTFFIND and CTFTILT or just CTFTILT. Normally, I do not like to process micrographs for the CTF, but we should probably implement something.

Why does one micrograph in a session have a different number of pixels than the others?

Actions #4

Updated by Dmitry Lyumkis almost 9 years ago

I have not been able to track down anything obvious for why one micrograph fails, but the others do not. Will keep looking. It seems to be specific for CTFTILT.

Actions #5

Updated by Neil Voss over 8 years ago

  • Status changed from New to Assigned
  • Assignee changed from Neil Voss to Dmitry Lyumkis
  • Priority changed from Urgent to Normal
Actions #6

Updated by Emily Greene over 8 years ago

  • Status changed from Assigned to Closed
  • Estimated time set to 1.00 h
  • Workaround updated (diff)

Changing the field size from 512 pixels (default) to 256 pixels is our workaround. We're not quite sure why, but it has worked on multiple datasets so it's not a huge problem.

Actions #7

Updated by Anchi Cheng almost 8 years ago

  • Related to Bug #4437: ctftilt not working added
Actions #8

Updated by Anchi Cheng almost 8 years ago

  • Related to deleted (Bug #4437: ctftilt not working)
Actions #9

Updated by Anchi Cheng almost 8 years ago

  • Has duplicate Bug #4437: ctftilt not working added
Actions

Also available in: Atom PDF