Three Dimensional Reconstruction
of Single Particle Specimens using
Reference Projections
This page describes a procedure for creating a 3D ribosome structure.
An automated particle selection process serves to identify ribosomes
from digitized electron micrographs. The image series is then aligned
relative to reference projections through shifts and rotations using
either AP MQ or AP MR command. A 3D iterative reconstruction is calculated. Difference map and its
significance can be calculated.
updated 1/3/97, Amy Heagle
updated 2/5/97, Paul
Automated Particle Selection
1.Adjust the micrograph dimensions
Use the Spider command FI to find the dimensions of the
micrograph. Check that the image dimensions can be interpolated
down to exactly 1/4 the original size and then verify that the
interpolated image size can be Fourier transformed (see list of
appropriate image sizes in the FT page of the index of spider
operations). If the original image dimensions need to be
changed, window the micrographs accordingly.
b01.win
2.Select a background noise file
Open one micrograph in WEB and use the Pixel utility to identify
coordinates for a window containing background noise
(no particles). Window this region from the micrograph.
b02.noi
3.Create a mask file
Create a circle with dimension and radius (generous) corresponding
to the particle size.
b03.mod
4.Run automatic particle selection
The micrographs are interpolated down by 1/4, fourier filtered
with a Gauss-low pass, and then peak searched over regions
corresponding to particle dimensions.
b04.mpk
5.Verify automatically selected particles
Sort through the output files to eliminate any non-particles.
This is accomplished in WEB with the Categorize command by
montaging the particle image files, manually clicking on each good
particle, and then saving the good file numbers in a doc file.
The doc file then needs to be adjusted so that the image numbers
are in ascending order with sequential key numbers using the
following program.
b05.ati
At this point either Multireference Alignment using AP MQ
(preferably) or Multireference Alignment using AP MR can be
used. Follow either track to Iterative 3D reconstruction.
Multireference Alignment using AP MQ
6.Create a selection doc file for 87 reference projections
Reference projections are views of ribosomes collected in
a previous project. This file contains a column of numbers
ranging from 1-87 which will be used to call the reference
projections used in the alignment of the particles in step 9.
b06.spl
7.Obtain reference projections from a reference volume
b07.pjq
Angles corresponding to reference projections are located in
angf
and were generated using delta(theta)=15.
9.Align the particles to the reference projections AP MQ
This is a multireference alignment of an image series.
b51.amq
11.Rotate particles according to alignment parameters
Any particles that correspond to reference projections 88-174
are also mirrored since projections 88-174 are mirror images
of the first 87.
b12.alm
12.Compare aligned particles to reference projections
Create a doc file which identifies particles into reference
projection groups and displays the correlation coefficient
describing the relative similiarity of the particle to the
reference projection.
Warning: correlation coefficient is not normalized.
b14.cla
Note: Instead, faster Fortran program can be used:
group.f
/net/penang/usr1/pawel/useful-fortran-programs/group.exe
The particles corresponding to each projection can be viewed in
WEB under the Montage from doc file selection to visually
associate a particular correlation coefficient to the image.
The greater the value, the more similiar the particle to its
reference projection. Identify a minimun correlation coefficient
that describes true particles as opposed to erroneously selected
particles.
13.Again compare particles to reference projections, this time using
particles above a specified correlation coefficient. Repeat
b12.cla, eliminating particles with low correlation coefficients.
Again, montage from each reference projection doc file the aligned
particle images. Select Compute Average to view an average of
the particles displayed.
14.Compute averages for all reference projections
It is sometimes convenient to view all averages together by
montaging them in WEB.
b13.avg
14a.In case of strong overrepresentation of some of the angular
directions the numbers of images per directions can be limited
to certain number:b55.eqp,
or keep half of the best per direction, but no more
than specified number:b56.eqp.
go to 3D reconstruction
Multireference Alignment using AP MR
6.Create a selection doc file for 174 reference projections
Reference projections are views of ribosomes collected in
a previous project. This file contains a column of numbers
ranging from 1-174 which will be used to call the reference
projections used in the alignment of the particles in step 9.
b06.sel
7.Obtain reference projections from a reference volume
b07.pjq
Angles corresponding to reference projections are located in
angf
and were generated using delta(theta)=15.
8.If the doc file containing the good particles is large,
it can be broken up into manageable parts so that step 9 can
be run on many machines.
b08.ord
9.Align the particles to the reference projections AP MR
This is a multireference alignment of an image series through
shifts and rotations.
b09.apr
10.Combine all resulting alignment files into one doc file
b10.dli
11.Rotate particles according to alignment parameters
Any particles that correspond to reference projections 88-174
are also mirrored since projections 88-174 are mirror images
of the first 87.
b11.alm
12.Compare aligned particles to reference projections
Create a doc file which identifies particles into reference
projection groups and displays the correlation coefficient
describing the relative similiarity of the particle to the
reference projection.
b12.cla
Note: Instead, faster Fortran program can be used:
groupa.f
/net/penang/usr1/pawel/useful-fortran-programs/groupa.exe
The particles corresponding to each projection can be viewed in
WEB under the Montage from doc file selection to visually
associate a particular correlation coefficient to the image.
The greater the value, the more similiar the particle to its
reference projection. Identify a minimun correlation coefficient
that describes true particles as opposed to erroneously selected
particles.
13.Again compare particles to reference projections, this time using
particles above a specified correlation coefficient. Repeat
b12.cla, eliminating particles with low correlation coefficients.
Again, montage from each reference projection doc file the aligned
particle images. Select Compute Average to view an average of
the particles displayed.
14.Compute averages for all reference projections
It is sometimes convenient to view all averages together by
montaging them in WEB.
b13.avg
Iterative 3D Reconstruction
15.After deciding on a correlation coefficient threshold, create a
selection doc file which refers to the particles to be used in
the 3D reconstruction.
b14.pap
15a. (APMQ) b15.pap
16.Create a doc file containing particle file numbers to be used in the
3D as well as reference angles for each particle
b15.ang
16a. (APMQ) b16.ang
17.Compute the 3D reconstruction
b16.bpr
18.Split select file used in 3D into two separate select files to be
used in the following two 3D reconstructions for comparative
purposes.
b17.ode
19.Compute the 3D reconstruction of half of the available particles.
b18.bpe
20.Compute the 3D reconstruction of the other half of the available
particles.
b19.bpo
21.Compare the two half volumes
b20.rff
22.In UNIX, use gnuplot to view the resulting curve
Plot 'doccomp' using 3:5 with lines
23.Using different correlation coefficients, create various volumes by
repeating steps 12 through 22.
24.3D projection alignment
Compute a projection of the final volume, calculate distances
between projections, and convert output to angular doc file.
Calculate new, refined 3D structure using centered projections
and the correced angles from the angular doc file.
b21.prj
Difference Maps
30.Create difference map
b27.dif