Hi Joshua,

 

(1)

_rlnFinalResolution represents the estimated resolution where the program will not be fully sure about the fitting.

The name of this keyword was only for old versions of Relion. Currently, Relion has changed this keyword and also Gctf is not going to use this keywords any longer.

This is only the estimation, not determination, and could be sometimes overstimated or underestimated due to abnormal background.

We cannot accurately determine the real information limit of a micrograph as the high frequency is really noisy.

Water rings and detector artifacts could sometimes affect the estimation severely (hope to solve these problems in future).

 

(2)

_rlnFigureOfMerit  is cross-correlation after background subtraction, weighting and normalization etc.

For micrographs with the same sample particle distribution and imaging conditions from the same session , this could roughly tell the overall quality of your micrographs.

However, it could not be used to compare micrographs quality with significantly different conditons, e.g. 10 vs 100 particles, with carbon layer vs pure ice, 20e vs 50e etc.

 

(3)

The equiphase averaging will help a lot to decide whether the CTF is realistic estimation, whether the resolution limit is overstimated or understimated and whether the micrograph itself might have problems.

You could easily tell which case it is just by eye.

There is a log file called the xxxx_EPA.log. It will tell you more information once something is abnormal.

 

My suggestion is to try multiple criteria. Each time you just throw away the obviously problematic micrographs by one criterion and then another.

Finally, if you really want to make a confident decision for each micrograph, it might be helpful to open the shrinked micrograph at the same time with the same order as the ctf files, so that you can judge in both spaces.

 

P.S. I would be quite interested to collect any special micrographs in this two cases:

(1) the micrograph looks nothing wrong and the FFT transform is also OK by eye, but CTF determination or some evaluation result is obviously wrong.

(2) the micrograph is obviously garbage as a result of abnormal imaging, but the CTF determination works well as if this is a good micrograph in all aspects.

    (Note, garbage here only means the imaging garbage,  sample problem is not considered as this special case)

 

Best wishes,

Kai

 

---

发件人：Joshua Lobo <[log in to unmask]>

发送时间：2017-03-31 22:36

主题：[ccpem] CTF Final Resolution and CTF FOM using GCTF

收件人："CCPEM"<[log in to unmask]UK>

抄送：

 

Hi CCPEM  

 

My question is regarding the _rlnFigureOfMerit and _rlnFinalResolution  . Is the rlnFinalResolution an estimate to where say GCTF could accurately determine the CTF value of a given Micrograph?(before the CTF gets close together at high frequency) .  

 

I am unclear on what the Ctf Figure of Merit is .Could someone please clarify that .  Would sorting on either of these parameters help in weeding out bad micrographs? 

 

 

 

Sincerely 

Joshua Lobo