Homogeneous refinement GSFSC underestimate

For awhile I have been refining particle clusters that are identified/‘purified’ during 3DVA which has been quite illuminating for my system (GSFSC 0.143 between 2.6 and 2.1A for each cluster refinement). Now, however, with largely the same procedure (the only difference being that I have downsampled my particle box size from 512 to 256 for my input particles) I am being given a clear GSFSC underestimate (see curves). My pixel size is 0.86 A/pxl. The only non-default parameters that I have in the job are:

  1. C5 symmetry is applied
  2. Initial lowpass resolution = 10A (it is a high contrast sample with high symmetry)
  3. Optimize per-particle defocus (ON)
  4. Optimize per-group CTF params (ON)
  5. Cache particle images on SSD (OFF)
    Any ideas what is going on here?

In this FSC plot the Nyquist frequency is around 3.5 Å, so the pixel size must be ~1.7 Å. You have to use more finely sampled particles to get a higher resolution.

That was my thinking too but I am not sure how this occurred. Do you think think this pixel size mistake may have occurred during my downsampling job (going from box size 512 to 256)?

I see, this was already answered by you and others here: Downsampled particles need full resolution

My mistake was not re-extracting from micrographs

1 Like