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