Hi Liz et al,
We recently did some tests where we recorded data sets on the same grid
with GIF-K2 +/- VPP. The defocus was set to constant -0.5 um with the
VPP, and a range -0.5 to ~-1.0 um without the VPP. All other parameters
equal. What we noticed was that a much lower fraction of micrographs had
good Thon ring fits beyond 4-5Å (in Ctffind4) with the VPP, as compared
to without the VPP.
One possible explanation that we liked is that fitting an additional
parameter (phase shift) requires more signal, and our relatively dilute
particles didn't provide this in all images. Sounds to me like you
observed a similar phenomenon, except that you also have the compounding
effect of increased mag i.e. less scattering material in the image. Or
what do you think?
Sorry that I'm not providing the solution to save your data set (other
than: record another data set with higher sample concentration and/or
lower mag in case you want to keep using the VPP) but I thought the
side-by-side comparison would be interesting to hear about.
Cheers,
Lars
On 2017-06-29 07:26, Liz Kellogg wrote:
> Hi all,
>
> I recently obtained a cryo dataset with a Krios/K2/VPP setup at high-mag
> (angstrom per pixel scaling of 0.66 Angstrom, 20k particles). My
> particle has a high degree of helical symmetry, and in the past datasets
> of this size (very similar to this one) have achieved better than 3.5
> Angstrom resolution, but in the past I haven’t used this high of mag or
> the VPP.
>
> This dataset (at higher mag and with the VPP) refines to 4 Angstrom
> resolution. Typically I image over open holes, and I wonder whether the
> increase in magnification (and hence smaller field of view) is making it
> harder if not impossible to correctly fit a CTF model for my
> micrographs. I see thon rings to about 12 Angstrom resolution (attached
> example FFT), but not much more than that. The fact that I don’t see
> thon rings as far as 3 Angstrom doesn’t usually bother me, because I
> find this is usually true with my other dataset imaged over open holes
> (which refined to 3.4 Angstrom).
>
> Depending on my CTF fitting runs (GCTF or CTFFIND4), I can get slight
> changes in the estimated resolution of my reconstruction (4.1 for GCTF
> and 3.9 for CTFFIND4). GCTF reports that the fit of the CTF is reliable
> to between 3.3 - 4 Angstrom resolution on average, depending on the
> resolution limit I use for fitting.
>
> I’m puzzled as to why this dataset is not refining to better than 4
> Angstrom resolution. What else could I consider? If it's a problem
> fitting the CTF, then is there any way to rescue this dataset?
>
> Thanks a lot
>
> Liz
>
--
Lars-Anders Carlson
Assistant Professor
Dept of Medical Biochemistry and Biophysics
Wallenberg Centre for Molecular Medicine
Umeå University
901 87 Umeå, Sweden
http://www.medchem.umu.se/english/research/principal-investigators/lars-anders-carlson/
|