Issue with CTF correction in FREALIGN reconstruction
Forums
Hi,
I have observed very different results between two different approaches for CTF correction of map reconstructions performed with FREALIGN v9.11 for one dataset that I'm working on:
1st approach: normal CTF correction inside FREALIGN. Defocus and astigmatism values specified in .par file, WGH = 0.07.
2nd approach: correct the stack of particles for the CTF prior to reconstruction in FREALIGN, then run reconstruction with WGH = -1.0 to prevent any further CTF correction. I have a SPARX script in Python to do that (either phase flipping or full correction with restoration of amplitudes).
For the 1st approach, the map seems to be missing a lot of medium-resolution features (high frequency noise is still present). For the 2nd approach, the map looks fine. I wonder why do they look so different if defocus, astigmatism and amplitude contrast values provided to FREALIGN and SPARX are the same? Does FREALIGN apply filters other than the optimal single-particle filter activated by FFILT?
I have inspected the values of the inputs passed to ctfapply.f at the moment of reconstruction and they are OK.
I work with tilted specimens and I have an estimate for the defocus values at each particle's position within the micrograph. Evidently, these estimates have errors that increase with the tilt angle, so I would like to refine the defocus on a per-particle basis to try to improve the map.
Thanks for any help in understanding and solving this issue.
One additional information:
One additional information: reconstruction with RELION yields the same results for the two approaches described above. So the problem lies within my dataset or CTF parameters determination. But it still intrigues me the fact that if the CTF correction is done prior to reconstruction, the map looks good.
It sounds to me as if Relion
In reply to One additional information: by rdrighetto
It sounds to me as if Relion performs CTF amplitude correction, even if the CTF is "switched off". When the CTF correction is disabled in Frealign, there is no correction whatsoever and therefore, amplitudes especially at lower resolution will be strongly attenuated. This may be what you see in your reconstructions.
My previous post might have
In reply to It sounds to me as if Relion by niko
My previous post might have been a bit confusing. By "same results" I mean that with Relion I observe the same behavior as with Frealign. I think it has to do with the particle densities normalization being affected in a different way by pre-correcting the CTF. Or perhaps I am missing some conceptual difference between correcting for the CTF in 2D and in 3D?
Nevertheless, the map does improve after refinement with normal CTF correction inside Frealign. I'm just concerned by the reconstruction results being considerably different using pre-corrected stacks.