Yes it is indeed related to the low-pass filtered display in the older Relion version. Somehow the dose-filtered image from unblur looks normal but that from MotionCor2 has some lines.
In the newest relion2 display, all images look normal (no lines).
Thank you all for the help!
Weiwei
________________________________________
From: Collaborative Computational Project in Electron cryo-Microscopy <[log in to unmask]> on behalf of Sjors Scheres <[log in to unmask]>
Sent: Friday, October 14, 2016 7:23 AM
To: [log in to unmask]
Subject: Re: [ccpem] Fw: [ccpem] MotionCor2
The display artefact was solved in the latest beta version of relion-2.0
(I hope).
S
> Hi Weiwei,
>
> We had similar issues. It's a display artefact. Open the micrograph with
> e2display.
>
> Cheers
>
> paolo
>
> 2016-10-14 2:27 GMT+02:00 Weiwei Wang <[log in to unmask]>:
>
>> Right it seem not likely related to motion correction. Though, since
>> the
>> '-Crop' option corrected the problem I think it is less likely to be due
>> to a rotation/flip mismatch of the gain reference.
>>
>>
>> I am attaching 3 screen shots (all low pass filtered at 20A):
>>
>>
>> The images in "rect-sqr.jpg" was acquired with the correctly
>> rotated/flipped gain reference. On the left is the original sized
>> rectangular image and on the right is the cropped square image (by using
>> '
>> -Crop 7420 7420' option).
>>
>>
>> The images in "rect-sqr-wrngr.jpg" was obtained with incorrectly
>> oriented
>> gain reference ( also rectangular on the left and square on the right).
>>
>>
>> The last image 'rect-unblur.jpg' is the rectangular image corrected
>> using
>> "unblur_openmp_7_17_15.exe" (
>> "mag_distortion_correct_openmp_8_18_15.exe" was
>> used to apply gain reference).
>>
>>
>> The vertical lines in the rectangular micrographs look different from
>> the
>> stripes resulting from a misaligned gain reference I think. An
>> interesting thing, if I do not use low-pass filtering for display
>> (relion, set scale 1.0 and low pass filtering 0), these lines seem to
>> disappear. When processed with
>> "mag_distortion_correct_openmp_8_18_15.exe"
>> and "unblur_openmp_7_17_15.exe", these lines seem not presence even with
>> low-pass filtering. Maybe there is some difference in how gain reference
>> is used or the low-pass filtering (fourier cropping?) process?
>>
>>
>> Thanks a lot!
>>
>>
>> Weiwei
>>
>>
>> ------------------------------
>> *From:* Shawn Zheng <[log in to unmask]>
>> *Sent:* Thursday, October 13, 2016 6:18 PM
>> *To:* Weiwei Wang
>> *Cc:* [log in to unmask]
>> *Subject:* Re: [ccpem] Fw: [ccpem] MotionCor2
>>
>> At this point we can conclude these lines are not caused by motion
>> correction. I just saw Zongli's email and think he is very likely
>> correct.
>> One more thing to try, generate uncorrected sum without gain correction.
>> This can be done by not specifying "-Gain". Again use "-Align 0" to
>> disable
>> motion correction. Compare this image and the gain reference to check
>> the
>> orientation.
>>
>> Best
>> Shawn
>>
>> On Thu, Oct 13, 2016 at 3:01 PM, Weiwei Wang
>> <[log in to unmask]
>> > wrote:
>>
>>>
>>> ------------------------------
>>> *From:* Weiwei Wang
>>> *Sent:* Thursday, October 13, 2016 5:59 PM
>>> *To:* Shawn Zheng
>>> *Subject:* Re: [ccpem] MotionCor2
>>>
>>>
>>>
>>>
>>> Hi Shawn,
>>>
>>>
>>> Both the uncorrected (-Align 0) image and the 20A low-pass filtered
>>> gain
>>> reference have the lines. Maybe it is simply due to
>>> low-pass filtering? Seems somehow the images are scaled a bit
>>> differently
>>> also? Thanks a lot!
>>>
>>>
>>> Best,
>>>
>>> Weiwei
>>>
>>>
>>> ------------------------------
>>> *From:* Shawn Zheng <[log in to unmask]>
>>> *Sent:* Thursday, October 13, 2016 3:41 PM
>>> *To:* Weiwei Wang
>>> *Subject:* Re: [ccpem] MotionCor2
>>>
>>> Try to generate uncorrected sum using -Align 0 and see if you can
>>> duplicate these lines. If yes, we may need to look into individual
>>> frames
>>> and the gain reference. Also, try low-pass filter the gain reference to
>>> 20A
>>> and see if you can see these lines too.
>>>
>>> If these tests cannot duplicate these lines, I then need your stacks
>>> and
>>> gain reference to check by myself.
>>>
>>> Best
>>> Shawn
>>>
>>> On Thu, Oct 13, 2016 at 12:16 PM, Weiwei Wang <
>>> [log in to unmask]> wrote:
>>>
>>>> Hi Shawn,
>>>>
>>>>
>>>>
>>>> The lines seems to be in the summed micrographs (instead of Fourier
>>>> transform). The gain reference was flipped / rotated before using
>>>> MotionCor2 and it seems to be OK in other programs. I am attaching an
>>>> screen shot (on the left is the rectangular image and on the right is
>>>> the
>>>> cropped square image, low pass filtered to 20A) that shows what we
>>>> see.
>>>> Thanks!
>>>>
>>>>
>>>> Weiwei
>>>>
>>>>
>>>> ------------------------------
>>>> *From:* Shawn Zheng <[log in to unmask]>
>>>> *Sent:* Thursday, October 13, 2016 1:35 PM
>>>> *To:* Weiwei Wang
>>>> *Cc:* [log in to unmask]
>>>> *Subject:* Re: [ccpem] MotionCor2
>>>>
>>>> Hi Weiwei and Tat,
>>>>
>>>> I assume the vertical lines are seen in the Fourier transform,
>>>> correct?
>>>> If this is the case, combined with Tat's workaround, I would think
>>>> these
>>>> lines are due to orientation mismatch between frames and gain
>>>> reference.
>>>> The new version of MotionCor2 allows to rotate and flip gain reference
>>>> before it is applied to frames. These options can be enabled by
>>>> "-FlipGain"
>>>> and "-RotGain". You can download from http://msg.ucsf.edu/em/so
>>>> ftware/index.html.
>>>>
>>>> Best
>>>> Shawn
>>>>
>>>> On Thu, Oct 13, 2016 at 9:38 AM, Weiwei Wang <
>>>> [log in to unmask]> wrote:
>>>>
>>>>> Hi Shawn,
>>>>>
>>>>>
>>>>> Thank you for making this great software!
>>>>>
>>>>>
>>>>> We are currently trying to use it for motion correction. There is one
>>>>> problem that we encountered.
>>>>>
>>>>>
>>>>> The dimension of our K2 sensor (super-resolution mode) is 7420 x
>>>>> 7676.
>>>>> If we use this original size, the dose filtered output (*DW.mrc, with
>>>>> 5x5
>>>>> patches) looks dim and has vertical lines (along the long dimension).
>>>>> However cropping the image into 7420 x 7420 using the -Crop option
>>>>> solves the problem.
>>>>>
>>>>>
>>>>> I wonder if square images are required? Thanks a lot!
>>>>>
>>>>>
>>>>> Best,
>>>>>
>>>>> Weiwei
>>>>> ------------------------------
>>>>> *From:* Collaborative Computational Project in Electron
>>>>> cryo-Microscopy <[log in to unmask]> on behalf of Shawn Zheng <
>>>>> [log in to unmask]>
>>>>> *Sent:* Thursday, October 13, 2016 11:23 AM
>>>>> *To:* [log in to unmask]
>>>>> *Subject:* [ccpem] MotionCor2
>>>>>
>>>>> Hi all,
>>>>>
>>>>> I am the author of MotionCor2. I will be happy to answer MotionCor2
>>>>> questions and take your suggestions,
>>>>>
>>>>> Best,
>>>>>
>>>>> Shawn
>>>>>
>>>>
>>>>
>>>
>>
>
--
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
|