Poor CTF fit due to ice

Submitted by Luca Pellegrini on Tue, 2020-11-17 09:54

Forums

ctffind & ctftilt

I am struggling to perform the CTF fitting due ice contamination. In a large proportion of my images, the CTF model fits the strong signal from the ice rather than the weaker Thon rings of the sample. I was wondering if there is a way to exclude the ice rings from the CTF correction (like in X-ray diffraction data)? I am using CTFFIND4 in Relion 3.1.

I have attached two examples of CTF fit, one where the fitting worked and one where it didn't.

Thank you for your help and best wishes,

Luca

Alexis Tue, 2020-11-17 13:04

Hi there,

There is no built-in feature to ignore the water ring. Two things I can suggest:

(1) Just fit to a lower resolution, say 4Å, so that the water ring is ignored

(2) Do the bad fits score lower? If so, only keep results from micrographs with high score (or maybe fit resolution?). Do you really want to keep the micrographs with so much ice contamination anyway?

Cheers,
Alexis

Luca Pellegrini Wed, 2020-11-18 06:47

Hi Alexis,

Thanks for getting back to me. I am running CTFFIND4 in Relion, with default min and max resolution values of 30A and 5A (as set in the Relion GUI). So this should in principle take care of the ice rings, as you suggest, but it doesn't seem to work. What would be the CTFFIND command line instructions to set the resolution range?

Best wishes,

Luca 

Alexis Wed, 2020-11-18 12:04

Hi Luca,

It looks like you may have told Relion (and therefore CTFFIND4) the wrong pixel size. In the diagnostic images output by CTFFIND4, the top left quadrant shows the theoretical CTF up to the radius corresponding to the max resolution for the fit. If indeed you told it to fit 30-5Å, it's drawing that 5Å radius way out beyond the ice ring, which suggests it got the wrong pixel size. Can you double check this?

Cheers,
Alexis

Luca Pellegrini Thu, 2020-11-19 10:00

Hi Alexis,

Yes, I got the pixel size wrong. It's a super-resolution dataset and I input twice the pixel size, which explains the anomalous apparent resolution of the model rings. I've imported/motion-corrected with the right pixel size, and now CTFFIND4 works nicely. Apologies for the trivial error and thanks for your help.

Best wishes,

Luca  

rbaxter Fri, 2021-08-27 16:36

Dear Alexis/Luca,

Newbie follow-up question. So if I just collected data at NSLS-II and they say the K3 mode was "counted super-resolution" with pixel size 0.825 Angstrom (105,000x mag) does that mean the actual pixel size is 0.4125 (8K) unless I use bin_factor 2 during Motion Correction to bring it to 0.825?

Because I imported the data into RELION with apix 0.825, ran Motion Correction with bin_factor 1, and then CTF correction gives terrible results like described here (MaxRes >100 A). But if I run CTFFIND separately and input pixel size 0.4125, then I get a nice fit (MaxRes 4.2 A).