Comparison between methods used to estimate and fit CTF, Warp versus CTFplotter in IMOD

[Cyan Ching]

Dear all,

Good morning/afternoon/evening!

Does anyone happen to have knowledge on the comparison between methods used to estimate and fit CTF in Warp versus the CTFplotter in IMOD? 

The latter uses the Welch method which divides image FT signal into overlapping snips and averages them after edge effect correction (tapering, the signal is averaged with a Gaussian-like function). With this approach, CTFs are smoothed at high spatial frequencies to rid of noise. 

Using Warp, however, it seems CTFs are generally well-preserved across spatial frequencies when expected max defocus is set as 5 in both Warp and CTFplotter. 

Now I am confused about the difference between CTF estimation/fitting methods implemented in Warp and CTFplotter as I read Warp also does the following: 'Like in the frame alignment procedure, groups of patches matching the desired PS size (for example 5122–1,0242 pixels) are extracted with a spatial overlap of 50% from the raw movie data, transformed into Fourier space and converted to PS by taking their squared amplitudes.'

Thank you for your help in advance!

Best,

Cyan

[teg...@gmail.com]

Hi Cyan,

I'm not familiar enough with IMOD's implementation to comment on the details, but Warp's algorithm is pretty close to your description of IMOD's. The tapering in real space shouldn't have a noticeable effect on the amplitude spectrum unless the defocus is very significant. One possible visual difference is that Warp rescales the 1D amplitude spectra from all patches to one common defocus value for display in the GUI, which is why you may see high-res rings better.

One possible advantage of Warp is that it performs additional regularization over the entire tilt series and fits the global sample inclination (useful for lamellae). Again, I don't know if IMOD does something similar as well. I'm not aware of any scripts for importing one program's results into the other, so if you want to compare the accuracy, you might need to come up with a custom solution.

Cheers,

Dimitry