Hi,
> Is there any documentation/guidance for evaluation of parameters? For example, for my data rlnCtfFigureOfMerit ranges from 0.02 to 0.15; such a figure of merit would be suboptimal during phase determination in crystallography. Is there a range of typical/desirable FOM here? Likewise for astigmatism, defocus angle and such?
Although some people reject "bad" images by various criteria,
I myself rarely find it useful or improve the resolution.
rlnCtfFigureOfMerit is a score reported by CTF estimation programs RELION calls
(CTFFIND4 or GCTF). Each of them implements different target functions and values are not
directly comparable. Moreover, micrographs containing high contrast contaminations (e.g. big ice
or carbon edge) tend to give stronger Thon rings and thus better score, but obviously this
is not what you want. Thus, this is not a good criterion to find "bad" images.
rlnCtfEstimatedResolution is also program dependent and you should not
worry about absolute values (i.e. 4 does not mean you cannot get a 3 A structure
from the micrograph). That being said, large values (> 10 A) usually means a failure
to fit and it is usually harmless to remove such images. If you have lots of failures,
it is worth checking errors in parameters (e.g. didn't you make a typo in the pixel size,
Ac or Cs? Is the search range adequate?).
Defocus and astigmatism: as long as you can fit them properly and use the
right box size, large values are harmless. If the defocus values are very
off from what you requested in EPU/SerialEM/Leginon etc, it is worth
checking data collection parameters (e.g. focusing interval, eucentric height).
Nonetheless, it is worth looking at logfile.pdf to detect problems during data collection.
Astigmatism usually drifts only smoothly over a data collection session.
CtfEstimatedResolution tends to vary with ice thickness. Since EPU starts
at the center of a grid square (thin ice) and goes to peripheries (thicker ice),
the values tend to get slightly worse until EPU moves to the next grid square.
Best regards,
Takanori Nakane
P.S.
> ranges from 0.02 to 0.15; such a figure of merit would be suboptimal during phase determination in crystallography.
Even such bad phases can sometimes be improved by iterative density modification
with auto-tracing. See Fig. 4 of my old paper: http://dx.doi.org/10.1107/S139900471501857X.
> Thanks,
> Ketan
>
> -----Original Message-----
> From: Takanori Nakane <[log in to unmask]>
> Sent: Tuesday, October 13, 2020 11:40 PM
> To: Gajiwala, Ketan S <[log in to unmask]>; [log in to unmask]
> Subject: [EXTERNAL] Re: [ccpem] Micrograph selection after motioncorr2
>
> Hi,
>
>> I am assuming that it might be advisable to remove micrographs that
>> have " too few valid local trajectories to fit local motion model” from further analysis.
>
> No, this is not necessary.
>
> In general, having a very small fraction of outliers does not make the resolution worse.
> Don't worry too much.
>
> Best regards,
>
> Takanori Nakane
>
>> Can someone please provide a pointer on how to achieve it, and if there are general suggestions for subset selection criteria?
>>
>> (I am working with a dataset of 1004 movies, which generally have
>> rlnAccumMotionTotal in 20’s and _rlnAccumMotionEarly of ~2-4, but
>> there are few outliers with values ranging from 35 – 2000.)
>>
>> Thanks,
>>
>> Ketan
>>
>>
>> ----------------------------------------------------------------------
>> ----------------------------------------------------------------------
>> ----------
>>
>> To unsubscribe from the CCPEM list, click the following link:
>> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
>>
>
>
> ########################################################################
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
>
> This message was issued to members of www.jiscmail.ac.uk/CCPEM, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
>
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
This message was issued to members of www.jiscmail.ac.uk/CCPEM, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|