Hi Craig,
In practice groups don't seem to matter too much, as long as there are
enough particles to make robust sigma-noise and intensity scale-factor
estimates. I wouldn't worry too much about the details here.
The number of groups asked for is indeed not exactly the number you get.
I wouldn't make groups on rlnNormCorrection (which is a per-particle
feature, affected by the presence of neighbours/contamination etc). Better
values to group on are rlnDefocusU/V or the intensity scale factor i the
data_model_groups table in the model.star file. On the Wiki there is an
(old) shell script that does just that.
HTH,
Sjors
> Hi Sjors,
>
> I was hoping you could answer some follow-up questions I had after reading
> your previous post on the related matter:
>
>> Hi Jian,
>> If you had regrouped the individual sections (using the Display GUI),
>> they
>> will all have group-names like group_1, group_2, etc.
>> You could rename the group-names of the second section group2_1,
>> group2_2,
>> etc in order to prevent overlapping groups from the different sections,
>> as
>> these are not necessarily similar in the refined scale-factor (which is
>> the main divider of different groups).
>> HTH,
>> S
>
> It isn't uncommon after some particle filtering for me to start getting
> some warnings about groups that are too small. I usually use the regroup
> feature when opening a *_model.star file. So I have two quick questions
> about the behavior of this option:
>
> 1) The number entered for number of groups in the UI is not usually what
> the CLI app reports as the number of groups written. Is this because the
> group count is modified to create groups of equal size?
>
> 2) What is your current recommendation for number of groups vs. number of
> particles per group? The docs list 20-100 particles in a couple of
> places, but the latest UI seems to limit the number of groups to <1000.
>
> 3) (Small feature request if I may) If the number of particles per group
> is more important, could the UI option be changed to set this instead?
>
> In the meantime I wrote a small script to parse and rewrite the particle
> STAR files but I did get confused about wether I should change
> `_rlnMicrographName` to regroup the particles, as mentioned in the CLI
> outputs, or wether I should change the `_rlnGroupNumber` and
> `_rlnGroupName` fields. I was also not sure what criterion to use- but
> your previous reply might suggest `_rlnNormCorrection`?
>
> My current strategy would be to sort the entire particle set by
> `_rlnNormCorrection` and then split the sorted list into groups of equals
> size N (~50-100). Does that sound like it would work well?
>
>
> Thanks!
> -Craig
>
--
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
|