fsc calculation

[cyanpku]

Hi Catsten,

After playing with the CTF correction, I do see an improvement of resolution near the centre of the tube: there's an a-hlx and beta-strand used to be one continuous rod, now they are clearly separated. However, the fsc estimation is the same for the two reconstructions. I'm wondering how does SPRING perform fsc calculation? Does it use the entire reconstruction or mask out one asymmetric unit? Can I specify to which radius to calculate the fsc?

PS is there any plan to include ctfffind4 for micctfdetermine?

Thanks a lot!!

Jing

[Carsten Sachse]

Dear Jing,

On Monday, April 25, 2016 at 11:51:31 AM UTC+2, cyanpku wrote:

Hi Catsten,

After playing with the CTF correction, I do see an improvement of resolution near the centre of the tube: there's an a-hlx and beta-strand used to be one continuous rod, now they are clearly separated. However, the fsc estimation is the same for the two reconstructions. I'm wondering how does SPRING perform fsc calculation? Does it use the entire reconstruction or mask out one asymmetric unit?

  The FSC is estimated on the entire 3-D reconstruction. In addition to a soft cylinder mask, I have implemented the structure-based mask deconvolution scheme by phase-randomization according to Chen et al. 2013 Ultramicroscopy, which you should be able to recognize by the dip of the FSC curve.

 

Can I specify to which radius to calculate the fsc?

The radius is specified at the beginning of your refinement run: Estimated helix inner and outer diameter in Angstrom 

PS is there any plan to include ctfffind4 for micctfdetermine?

 Yes, there are concrete plans. We are currently testing it locally. In the next release, it will work with ctffind4. I will let you know.

Best wishes,

Carsten 

Thanks a lot!!

Jing

[cyanpku]

Hi Carsten,

Thank you very much for the clarification! How can I turn on the structure-mask fsc? In the fsc output file, I only have the unmasked and the cylinder masked data ....

Jing

[Carsten Sachse]

Dear Jing,

The structure-masked FSC will only be plotted for resolution smaller than 12 Angstrom. Below that I believe does not make much difference and also doesn't bear much meaning. This feature was introduced with SPRING v0.83 and is now available with SPRING v0.84

Best wishes,

Carsten

[cyanpku]

Hi Carsten,
Thanks a lot for your reply!
It's still puzzling why the structure masking is off for me : (
My emspring version is v084_1470.
Reading from the report.log, the unmasked FSC and cylindral masked FSC for my reconstruction are:
LR 8.4apix: (16.8, 16.8)
MR 3.6apix: (10.4, 8.1)
HR 2.4apix: (10.2, 7.8)
MaxR 1.2apix: (10.2, 7.0)
The structure mask is off all the time although the resolution (FSC0.143) improves better than 12A after the MR iterations. The resolution at FSC 0.5 is higher than 12A as well. Which fsc criteria does spring use to decide switching on the structure masking? Or is it discontented about my reconstruction didn't hit the Nyquist frequency at the end of MR refinement?
Thanks a lot for your help!
Jing

[Carsten Sachse]

Dear Jing,

Okay, it's a bit more complicated: if the 0.75 cut off is beyond 1/12 Angstrom (this was the arbitrary cutoff I chose), it will turn on the structure masking for FSC computation. Do have a funny looking FSC curve? Otherwise it should work. Again, I am not convinced that that that resolution, i.e. 9 - 12 Angstrom, this will make a meaningful difference. Remember, a good 10 Å map will give you alpha-helices as tubes...

 I hope this makes sense.

Best wishes,

Carsten  

[cyanpku]

Hi Carsten,

Attached is the screenshot of the diagnose file for the last iteration of segmentrefind3d, the fsc shows a few sharp spikes near the 10A, does it look normal? Our in-house microscope has been broken since I collected the last dataset (on a grid with thick ice) two months ago. I'm trying to squeeze the last drop of information out of the not-so-good dataset while waiting for the scope to recover, but maybe it's time to stop messing around spring until I get better data ...

Thanks a lot for your help!

Jing