Classify particles corresponding to each reference projection

verify-class-byview.spi reads: views/prj###/sel_proj , views/prj###/sel_part_byv and views/prj###/stkfilt .

Verify particles

For each each reference projection, the aligned particles will have been assigned to classes, and sorted by cross-correlation coefficient. Ideally, the good particles will fall into discrete classes, and non-particles into different discrete classes. Outliers can be added or removed individually.

Different levels of control over the particles kept are offered. Recommended is to open each montage of particles belonging to each class, and select good particles. The fastest method is to select entire classes of particles. In between the two, in terms of control, would be to manually select only the particles with the highest CCC values.

1. Use: verifybyview.py -- Documentation.
   The simplest way to start it would be to, from the Averages/ directory, type:
   verifybyview
   Adjust the parameters in the initial popup window, click Continue, and a montage of class windows will appear. Prior installation of SPIRE is required.
2. WEB -- Documentation

Verify particles in each of the reference view and don't forget to save views/prj###/sel_class_good.

Combine particles from good classes

[Optional] Re-screen the particles by view

For an additional layer of quality control, the newly created lists of good and bad particles by class can be re-screened for outliers.

To look for bad particles in the selection file, in the directory: Averages/Views, type (with the appropriate extension):
montagefromdoc.py   prj001/sel_part_byv_bad.dat   prj001/avgfilt.dat
There should be a settings file, .montagefromdoc, present with reasonable settings. Select only the bad particles, and write the selection file to: views/prj###/sel_part_byv_notgood.

Conversely, to look for good particles among the de-selected particles, in the directory:Averages/Views, type (with the appropriate extension):
montagefromdoc.py   prj001/sel_part_byv_good.dat   prj001/avgfilt.dat
Select only the good particles, and write the selection to: views/prj###/sel_part_byv_notbad.

Both of the above two steps are optional, i.e., you can perform one or the other, or both. If you run either one, you'll need to run the following procedure.

If this step is performed, remember in later steps to substitute views/prj###/sel_part_byv_goodB where appropriate.

[Optional] Check the spatial distribution of views (angular directions) for the reconstruction

The following two procedures are same as those run previously, except for one input setting used. Namely, change the parameter [wantGoodYN] to 1 to use the verified particles.

plot-ref-views.spi reads: views/parts_vsview_good. It creates a script file with gnuplot commands:

¤

parts_vsview_good.gnu

Gnuplot script to plot histogram of number of particles vs projection view.

The output Gnuplot script parts_vsview_good.gnu will plot number of particles vs view after verification. The histogram plot should be displayed automatically, if not click here.

.

The output Gnuplot script parts_vsview_good.gnu will plot number of particles vs viewafter verification. The histogram plot should be displayed automatically, if not click here.

show-ref-views.spi reads: ../Alignment/refangles, and views/part_vsview_good. Change the parameter [wantGoodYN] to 1 to use the verified particles. It creates:

¤

show_ref_views_good:

SPIDER image showing distribution of sample images among the various reference projections.

This image shows the various angular reference groups represented by small circles. The radii of these circles are proportional to the number of particles in each group. Display the image in: WEB or qview.py.

.

[Optional] Compare number of particles before and after verification as a function of defocus level.

With micrographs from all defocus values combined, it is possible that close-to-focus images may be discriminated against, for example because of lower CC values or simply being harder to see.

The output Gnuplot script good_parts_vsmic.gnu will overlay the number of particles before and after verification. The histogram plot should be displayed automatically, if not click here. See example on right:

Note that, depending on the origin of the ../Particles/good/sel_part_**** files, the quality of the "initial" particles may or may not be meaningful. For example, if no screening was done following lfc_pick.spi, there may be a large excess of non-particles in the ../Particles/good/sel_part_**** files.

Now, resume the "normal" flow of operations, starting with best-imgs.spi. Remember to set it to select the "good" particles.

---

References

1. Shaikh TR, Trujillo R, LeBarron JS, Baxter WT, Frank J. (2008)    Particle-verification for single-particle, reference-based-reconstruction using multivariate data analysis and classification. Journal of Structural Biology, Volume 164, Pages 41-48.

Page updated: 2016/10/19     Author: Tapu Shaikh