[3dem] [ccpem] lost gain reference image (Camera Normalization)

[Marin van Heel]

Dear Oliver

Unfortunately, that is not feasible in general. Our claim is that using 
the "/a posteriori/ camera correction" you will end up with a 
better/larger set of particles simply because there is less "fixed 
pattern noise"  in the data (and also less fixed pattern gain noise.) In 
other words the "with" versus the "without" picking coordinates would be 
different, the number of picked particles should be larger, the number 
of rejected particles further down the pipeline should be smaller, and 
the rotation/shift/ movie alignment parameters different.  If one would 
apply "camera-corrected picking coordinates" to not corrected images, 
one would improve the "not corrected" results beyond the "fully 
not-corrected" processing results. These would still be a bit worse than 
the original fully-corrected processing results.  If, on the other hand, 
one would apply the "not corrected" co-ordinates to the "corrected" 
images, those would do slightly better than the
"fully not-corrected" processing results due to the removal of the fixed 
pattern noise. However, one would miss out on the major "fully 
corrected" processing advantages: more particles and better movie 
alignments.  Are you still with me?

Bottom line: use the FRC as an immediate proof of the camera correction 
and stop thinking about the issue henceforth. ;)

By the way, the camera correction even improves the power spectrum and 
thus the CTF determination...  But that we will show in a forthcoming 
publication!

Cheers,

Marin


On 02/10/2018 13:03, Oliver Clarke wrote:
> Hi Marin,
>
> Couldn’t you just extract the same final particle set from both sets 
> of micrographs (with/without a posteriori gain correction) and refine, 
> though, without going through the whole pipeline independently for both?
>
> Knowing that it affects the quality of the final density map would 
> definitely be helpful, I think.
>
> Cheers
> Oli
>
> On Oct 2, 2018, at 11:51 AM, Marin van Heel 
> <0000057a89ab08a1-dmarc-request at JISCMAIL.AC.UK 
> <mailto:0000057a89ab08a1-dmarc-request at JISCMAIL.AC.UK>> wrote:
>
>>
>> Dear Dimitry,
>>
>> We have shown that you can perform movie alignments on a more local 
>> basis without using very extreme low-pass filters (sometimes 
>> described as "B-factors"). Thus you will necessarily have a larger 
>> yield of usable particles from the same set of micrographs. That is 
>> more than sufficient evidence of improvement!  The FRC is a metric 
>> that is local to your correction operation and that measures the 
>> improvement directly. The final 3D map resolution only comes at the 
>> end  of a long pipeline, that any two people will perform differently 
>> and that is too indirectly related to the very early data-set 
>> correction. Bottom line: the FRC metric is necessary and sufficient 
>> to show the data-set improvement by the camera correction. However it 
>> does not necessarily and sufficiently guarantee that nobody will 
>> generate gold-standard garbage further down the pipeline. ;)
>> Cheers,
>> Marin
>>
>> On 02/10/2018 07:54, Dimitry Tegunov wrote:
>>> Dear Marin,
>>>
>>> do you have results showing that the proposed correction improves 
>>> the final map resolution vs. conventionally gain-corrected movies? I 
>>> think the FRC curves are necessary and sufficient proof , but not 
>>> sufficient to prove the advantage of your approach.
>>>
>>> Cheers,
>>> Dimitry
>>>
>>> On Mon, Oct 1, 2018 at 9:23 PM Marin van Heel 
>>> <0000057a89ab08a1-dmarc-request at jiscmail.ac.uk 
>>> <mailto:0000057a89ab08a1-dmarc-request at jiscmail.ac.uk>> wrote:
>>>
>>>     Dear Da,
>>>
>>>     In IMAGIC-4D  you can perform the necessary camera correction!
>>>     (https://www.nature.com/articles/srep10317). It does it better
>>>     than any
>>>     manufactures correction and improves the data significantly even
>>>     when
>>>     performed after using the standard gain correction.
>>>
>>>     Cheers,
>>>
>>>     Marin
>>>
>>>
>>>     =====================================================
>>>
>>>     On 01/10/2018 15:36, Da Cui wrote:
>>>     > Hi all,
>>>     >     The gain reference image for one dataset was missing by
>>>     accident. In order to achieve a more accurate motioncor result,
>>>     does anyone have idea about how to generate a gain reference
>>>     image from the dataset (around 3k movies)?
>>>     >     Thank you so much for your help!!!
>>>     > ---Da
>>>     >
>>>     >
>>>     ########################################################################
>>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20181003/b5da1ff7/attachment-0001.html>