question about single particle ice images
Forums
We've run into some difficulties with using CTFTILT (and CTFFIND) with cryoEM images of single particles in ice. After CTF determination, CTFTILT outputs an image with the fitted CTF on the left and the power spectrum on the right and it ought to be possible to confirm that the CTF has been fitted correctly by seeing if the Thon rings in the two halves of the images match up. Unfortunately, the Thon rings appear to be invisible in the case of an entire dataset of single particle ice images. (This does seem to work fine with many of the images in another dataset I have from 2D crystals on carbon.) I've experimented with different tile sizes and values of the pixel averaging parameter (PAve); but the Thon rings still seem to be invisible for images in this single particle ice dataset.
I notice that there only appears to be a figure with an image comparing the fit CTF and the power spectrum in the CTFTILT paper (Mindel and Grigorieff, 2003, JSB 142, 334-347) but not for single particle ice data although there is quite a lot of other results concerning single particle ice data in the paper. Is this a useful diagnostic to use for single particle ice data? Also, I was wondering about using the tilt angle results as a diagnostic instead. It seems that if CTFTILT finds a tilt angle very far from the expected result (e.g. a high tilt angle when the images are nominally untilted) then the defocus values are unreliable. However, I'm not sure how much the defocus values can be trusted in the case of a difficult image, if a sensible tilt angle is found? If the Thon rings are invisible in the power spectrum there still isn't a direct way to check the fit provided by the defocus values. (I noticed that that I would tend to get sensible tilt angles in the case of the difficult images if I used a value for PAve of 1 rather than 2. As this was with untilted images, I wondered if there was some bias favouring a low tilt angle rather than tending to get a better fit when I use a PAve of 1 rather than 2),
William
The diagnostic Thon ring
The diagnostic Thon ring image written out by CTFTILT is calculated from only a small part of the image that consists of a narrow strip running parallel to the tilt axis. This is done to average only power spectra that correspond (more or less) to a single defocus value. If no carbon is present in the image, it is usually difficult to detect Thon rings. The actual CTF fit is done across the entire image, however, and often gives reasonable results even if the diagnostic image does not show any Thon rings. The latest CTFTILT version allows the user to provide an expected tilt value to make the algorithm more robust (earlier versions of CTFTILT that do not ask for an expected tilt value should not have a bias for any tilt values). If CTFTILT produces tilt values that differ significantly from the expected value this is likely an indication that it did not latch onto the correct defocus and/or tilt values. One way to make CTFTILT more robust in cases where the tilt is expected to be small is to run CTFFIND3 first and get a reasonably good average defocus value for the image. The defocus search range in CTFTILT can then be made much smaller to include only values close to the value found by CTFFIND3.