[3dem] 2D classification resolution problem
Takanori Nakane
tnakane.protein at osaka-u.ac.jp
Fri Sep 8 01:40:29 PDT 2023
Hi,
Please don't drop CC to 3dem. I am a computational scientist
and don't know much about sample preparation. You have a better
chance of getting useful replies on the mailing list.
> orientation bias is enormous
You should also look at directional FSC.
It is what really matters.
> not even for the 0° data set
This means that your new grid is somehow worse, e.g. thicker ice.
> do oyu think the global vs local CTF estimation has
> on the 2D classification?
I recommend that you proceed to 3D refinement and judge the quality
by resolutions estimated from gold-standard FSC.
Best regards,
Takanori Nakane
On 2023/09/08 17:34, Dario Saczko-Brack wrote:
> Hi Takanori,
>
> thanks for your reply in the group. Actually the tilted data set was my
> last resort :-D I got the very best grids when using Au-Flat (or
> UltrAuFoil) grids, where my protein was much higher concentrated while
> not being aggregated (as it was the case for C-Flat). However, the
> orientation bias is enormous (I attached a .bild file), and I just
> couldnt get rid of it.
>
> So I tried 0° and 30° but now I have the problem that I cannot reproduce
> the previous 2D classification quality, not even for the 0° data set.
>
> Anyhow, what are your suggestions to overcome the orientation bias and
> how big of an effect do oyu think the global vs local CTF estimation has
> on the 2D classification?
>
> Best
> Dario
>
> On 07.09.23 11:57, Takanori Nakane wrote:
>> Hi,
>>
>> Tilted data collection means that defoci vary a lot within a micrograph.
>> RELION uses CTFFIND, which finds only one defocus value per micrograph.
>> If the dataset is good enough, you can repeat CtfRefine and Refine3D
>> to bootstrap (see Fig 3 of
>> https://urldefense.com/v3/__https://elifesciences.org/articles/42166__;!!Mih3wA!AnimFSo7qGvWrVPMUf3pJDzBeoDKRfRjvihyb6Diktr0NQutZuEqaIb6gXZVx8uSCtu5m9NBsxsf1BiZCFr5NAC1H3AmRizEXqA$ ). Otherwise, you might need local CTF estimation by other programs such
>> as Warp.
>>
>> Even if you manage to get local CTF, a tilted grid means a longer
>> path electrons have to pass through; so this has the same effect
>> as thicker ice.
>>
>> In my opinion, tilted data collection is the last resort against
>> preferred orientation problems. If you can resolve the issue by
>> other means (e.g. optimization of blotting parameters, additives,
>> slightly different construct), that would be much better.
>>
>> Best regards,
>>
>> Takanori Nakane
>>
>> On 9/7/23 18:48, Dario Saczko-Brack wrote:
>>> Dear all,
>>>
>>> I have acquired a small cryo EM data set of a protein (ca 120kDa) and
>>> processed it with Relion and Cryosparc. Since the 2D classes looked
>>> promising, but the orientation bias was very strong and the number of
>>> usable holes was very small, I made new grids and acquired another,
>>> much larger data set of the same protein at 0° and 30°.
>>>
>>> For some reason, I dont get to a similar resolution as the first data
>>> set. Any idea why? Would you say that is rather due to not optimal
>>> processing parameters or is it due to bad ice quality? Meaning, is it
>>> worth to invest more time into processing or into making new grids?
>>>
>>> Best
>>>
>>>
>>> Dario
>>>
>>> _______________________________________________
>>> 3dem mailing list
>>> 3dem at ncmir.ucsd.edu
>>> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
>> _______________________________________________
>> 3dem mailing list
>> 3dem at ncmir.ucsd.edu
>> https://mail.ncmir.ucsd.edu/mailman/listinfo/3dem
More information about the 3dem
mailing list