Instructions for processing Single particle crystals

1 Setup directories:

   mkdir doc
   mkdir patch
   mkdir coran
   mkdir classavg

2 Convert from mrc to spider file format with em2em

3 Invert the density with the spider command neg:

   
  spider wjr/spd
   neg
   <enter name of input file>
   <enter name of output file>
   en                 ; quit

4 Setup directories:

   mkdir doc
   mkdir patch
   mkdir coran
   mkdir classavg

5 Make a docfile that covers the entire area

   spider wjr/spd @make_grid

6 Optionally, low-pass filter the input file:

   spider wjr/spd
   fq
   <enter name of inpptu file>
   <enter name of output file>
    3                       (Gaussian lowpass filter type)
   0.2                      (Gaussian lowpass filter radius. 0.5=Nyquist. Use appropriate value)
   en            ; quit

7 Window the files:

   spider wjr/spd @window_grid

8 Do principal component analysis of the windowed files:

   spider wjr/spd
   ca s
   patch/patch****      <name of input files>
   1-1024               <particle numbers - use all>
   *                    <mask file - no mask>
   20                   <number of factors to calculate. Calculate many, don't need to use all>
   i                    <iterative pca>
   coran/coran          <output file prefix>
   en                 ; quit

9 Check Eigenvectors:

   gs EIGENVALUES.eps

10 Calculate eigenimages:

    spider wjr/spd @make_eigenimages

11 View these in Web, using montage command.

12 With a combination of viewing the images and values, choose the vectors you want to use

13 Do Hierarchical clustering using these vectors:

    spider wjr/spd
    cl hc
    coran/coran_IMC              [ input file]
    1-10                          [enter vector numbers determined above]
    0                             [equal weight for all factors]
    5                             [Ward's method of clustering]
    N                             [no postscript plot]
    Y                             [make a dendogram doc file]
    docdendro                     [output dendogram>]
    en                                ; end

14 Using Web, view the dendogram ("dendogram" command) and determine an appropriate cutoff

15 Make class files for the desired cutoff

    spider wjr/spd
    cl he
    0.3               [cutoff level determined above]
    docdendro         [dendogram doc file]
    doc/seldoc_**     [output class files]

16 Make average, variance files for the above classes and also calculate frc for each (estimate resolution)

    spider wjr/spd @frc_classes

17 Choose the best class as a reference for searching the original file

18 Search the raw file for correlation peaks with this best class.

    spider wjr/spd @corav1

19 Window out these peaks:

• Edit the script window2.wjr
• The script will automatically eliminate boxes that overlap the edge of the original image.

20 Do principal component analysis of these peaks

• Same as step 8, but enter newly windowed files for the input, and enter doc/goodpartpeak as the input selection docfile (rather than simply numbers)
• Similarly, repeat steps 9-16 for the new set of images.

21 View the chosen areas

    drawallboxes.wjr

• Set ALLOWTOPICVIEW =

-- BillRice - 06 Feb 2008