[3dem] 3dem Digest, Vol 142, Issue 68

Sun Jun 30 09:06:20 PDT 2019

Hi, Han,

85 % is a rough estimate and not an exact number, nor any of the number
in my explanation.  They are logical reasoning from looking at theoretical and
experimental DQE and FSC curves I have seen.

The basic idea is that for any FSC to pass the reviewer or more importantly
the author him/herself,  You will never claim the resolution at the half-map
Nyquist limit, right ?  We need to make sure that the curve drops to zero, and stay
there.  Beam tilt or other image aberration, badly masked map, or other bad
practice such as wrong MTF applied to the process may make it float above
or even dip to zero and come back up at higher frequency.
FSC is also not a step function, so there is always a non-zero
tail beyond your threshold of reported resolution.  In addition, there will be noise.
One single point at zero in a real experiment does not make it true that it is all at zero at
frequencies higher than that.

Therefore, I leave a bit of room for the evolution of the correlation being zero.  The actual
value of course depends on the detector behavior, your data size (for statistics), mask, and every
other things.  Some people will also be more confident in their “prediction” that FSC is
zero beyond their shorter stretch of frequency range.  I gave 85% as a conservative
estimate for K2 and as the point where the user should start to check it more carefully,
even though some of our own apoferritin data did get a reliable flat at zero-stretch
and good structure to higher Nyquist percentage (Large data set, well known mask, and
well behaved golden-standard FSC).  It is dangerous to give an ideal case as the recommendation to
general audience.

For super-resolution data, if the result resolution is so close to Nyquist of your maps,
and becomes difficult to get a good stretch of flat at zero-evalution, I would
recommend redo Fourier binning a little higher, not to the extent of unbinning. Even better to do
is to recollect the data with higher magnification (i.e. smaller non-super-res pixel size) so that you move
your structure signal to lower spatial frequency of the camera that has higher DQE.  More
work, but worth it.  After all, it is a rare opportunity to claim your favorite and non-ideal
biological specimen to be beyond 2.3-2.2 Å resolution.

One caution note, I was using the shape of K2/K3 DQE for my explanation, not Falcon3EC.
Falcon3EC DQE is quite different, although since it does not have super-resolution mode, it
does not fit the original question of “can I un-bin my super-resolution to get higher resolution structure ?”

Best,

Anchi

> On Jun 29, 2019, at 12:00 PM, 3dem-request at ncmir.ucsd.edu wrote:
> 
> Send 3dem mailing list submissions to
> 	3dem at ncmir.ucsd.edu
> 
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
> or, via email, send a message with subject or body 'help' to
> 	3dem-request at ncmir.ucsd.edu
> 
> You can reach the person managing the list at
> 	3dem-owner at ncmir.ucsd.edu
> 
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of 3dem digest..."
> 
> 
> Today's Topics:
> 
>   1. Re: 3dem Digest, Vol 142, Issue 60 (=?utf-8?B?66WY67KU7ZWc?=)
> 
> 
> ----------------------------------------------------------------------
> 
> Message: 1
> Date: Sat, 29 Jun 2019 20:15:00 +0900
> From: "=?utf-8?B?66WY67KU7ZWc?=" <psryubh at gmail.com>
> To: Anchi Cheng <acheng at nysbc.org>
> Cc: "3dem at ncmir.ucsd.edu" <3dem at ncmir.ucsd.edu>
> Subject: Re: [3dem] 3dem Digest, Vol 142, Issue 60
> Message-ID: <9199340E-44AD-4F30-A11C-12B3883B9B85 at gmail.com>
> Content-Type: text/plain; charset="utf-8"
> 
> Hi, Anchi
> 
> Cool explanation!
> 
> And here?s a question to your comment.
> 
> How the borderline; 85% nyquist has been determined and what limits that 15%?
> DQE or some other limitations in detector?
> And the value 15% is empirical or theoretical?
> 
> 
> Best regards,
> Han
> 
> 
> Bumhan Ryu
> 
> Postdoctoral Researcher
> Korea Basic Science Institute (KBSI)
> 161 Yeongudanji-ro, Ochang-eup, Cheongwon-gu, Cheongju-si
> Chungcheongbuk-do, Republic of Korea
> 
> Office) 043-240-5329
> 
> 
> 
> 2019. 6. 28. ?? 12:24, Anchi Cheng <acheng at nysbc.org> ??:
> 
>> Hi, Mohamed,
>> 
>> Just want to answer the first part of your first question.
>> 
>> Even though the signal you get in super-resolution mode is real, it is very dampened by DQE
>> of the detector.  Let?s say you have a signal count of 100 above the background across all frequency.
>> If DQE at your detector pixel limit is 0.2, and at 0.8 at zero frequency.  The output of that signal to
>> your image becomes 100*0.2 = 20 in the former and 80 in the latter.  This only gets worse as
>> you go into super-resolution frequency.
>> 
>> Since protein or other biomolecule do not give strong signal at high resolution to start with,
>> my ?signal across all frequency? assumption above is already false.  You can see that information
>> get through to your image at high frequency is much lower in reality.
>> 
>> If your 1? pixel data does not give reconstruction resolution to 85% Nyquist resolution (2.35 ? resolution
>> for 1 ? pixel) with all the modern single-particle processing tricks, the problem is in the specimen.  The very
>> dampened signal in the super resolution range by unbinning your images is not going to help you.
>> 
>>> 1-  Will my resolution improve if I do not apply bining? 
>> 
>> Best,
>> 
>> Anchi Cheng
>> NYSBC
>> _______________________________________________
>> 3dem mailing list
>> 3dem at ncmir.ucsd.edu
>> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <http://mail.ncmir.ucsd.edu/pipermail/3dem/attachments/20190629/c2649501/attachment-0001.html>
> 
> ------------------------------
> 
> Subject: Digest Footer
> 
> _______________________________________________
> 3dem mailing list
> 3dem at ncmir.ucsd.edu
> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
> 
> 
> ------------------------------
> 
> End of 3dem Digest, Vol 142, Issue 68
> *************************************