Hi Bo,
> 1. I wonder if there are ways to manually selecting certain resolution
> range for comparison purposes in fourier space during 3D refinement
> process. If so, how could I do it?
>
This does not exist. All shells are "chi2-weighted", i.e. you divide
squared differences by their sigma2.
>
> 2. I have a structure with multiple conformations and 8-fold
> pseudo-symmetry and 2-fold pseudo symmetry in another axis perpendicular
> to first axis. During first round of refinement, I imposed C8 symmetry.
> After convergence, I checked the #14 (_rlnMaxValueProbDistribution) which
> represents the likelihood value to associate particles with specific
> orientation and subset. For a large portion of particles in subsets that
> generates reasonable refined maps, on the other hand, the values are
> surprisedly low (<0.15) ( with c8 symmetry imposed). That leads to me to
> wonder the optimization process for fitting delta function from pdf
> function. I wonder if multiple peaks in pdf function are selected, and if
> so, what are the criteria to select those peaks? how is delta function
> optimized from pdf function? Sorry I does not find the detailed info in
> the documents or paper specifying this process from the material I have.
> Probably I missed the paper or document that specifically explain this
> process. Could you tell which paper or doc I should look at?
>
This process of "fitting a delta function" never happens in the
refinement. You always marginalize over all orientations and classes. The
output in the data.star files only contains the most likely orientation
and class, but this has nothing to do with the actual refinement.
Low maxima of the P-distribution indicate uncertainties in the angular
assignments, often due to low SNR in the images (for example making it
hard to distinguish between your pseudo 2-fold)
>
> 3. I would like to check the likelihood that one particle from raw image
> is associated with specific refined map from multiple refined maps,
> regardless of orientations. That may be useful as another criteria to
> determine whether one particle should go to specific subset or not. But I
> did not see that value in output files. Could you tell if there is a way
> to output that values in multiple model refinement process?
>
Sorry, also this is not possible. But then, it's all open-source: you
could go in and change it. :-) But be aware that the arrays with all
weights can e very long...
HTH,
S
>
> Have a good day and thanks in advance for suggestions!
>
>
> Bo Chen
>
>
> GS in BCM, USA
>
--
Sjors Scheres
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge Biomedical Campus
Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 267061
http://www2.mrc-lmb.cam.ac.uk/groups/scheres
|