Name

sxfind_struct - use the "common line" algorithm to assign initial values of phi, theta, psi to 2D average projections.

Usage

Usage in command lines:

sxfind_struct.py prj_stack outdir --ir --ou --delta --dpsi --lf --hf --rand_seed --maxit --debug --noweights --trials --given --first_zero --weights --MPIGA --pcross --pmut --maxgen --MPI --trials

Usage in python programing:

sxfind_struct(stack, out_dir, ir, ou, delta, dpsi, lf, hf, rand_seed, maxit, given = False, first_zero = False, flag_weights = False, debug = False, trials = 1)

Example:

sxfind_struct.py ave_15.hdf voronoi --ou=26 --delta=8 --lf=0.05 --hf=0.25 --rand_seed=10 --maxit=30

Note: the 2D input images have to be centered. The program does not determine translation parameters, it only finds the Eulerian angles.

Input

prj_stack

projection stack file

outdir

output directory

The parameters preceded with -- are optional and default values are given in parenthesis.

ir

inner radius used to the circle mask applied to the projections (default 1)

ou

outer radius used to the circle mask applied to the projections, ou is also used to set the length of 1D line projection =2*ou+1 (default ou=nx/2-1, where nx is the width of 2D image projection)

delta

accuracy of the angular search. Delta accuracy will define a set of quasi-evenly distributed projections directions to be used in the program (distribution 'P' of even_angles)

dpsi

accuracy of sinogram, dpsi refert to the angle step used to create the sinograms from the projections (default 1)

lf

for each 1D line projection only frequencies between minimum and maximum will be used, lf define the low frequency (default 0.0)

hf

for each 1D line projection only frequencies between minimum and maximum will be used, hf define the high frequency (default 0.5)

random_seed

the seed used for generating random numbers (set to random)

maxit

maximum number of iterations, each iteration will refine all angles for all projections (default 20)

trials

Number of trials for the MPI version

MPIGA

MPI version (Genetic algorithm)

pcross

Cross-over probability (set to 0.95)

pmut

Mutation probability (set to 0.05)

maxgen

Maximum number of generations (set to 10)

given

if present, set the initial angles of projections according the angles included in the header of each projection as objects xform.projection (default False)

first_zero

if present, set to zero the angles of the first projection, and keep them as zeroes for all iterations (default False)

noweights

do not use Voronoi weights (default False, i.e., the weights will be used)

Output

outdir/angles_000

this file contains Eulerian angles fount in trial #000 (if there were more trials requested, there will be more files numbered 001, 002, and so on)

outdir/plot_agls_000.hdf

this image in the hdf format contains visualization of the distribution of projections found during trial #000 (see also sxplot_projs_distrib)

outdir/structure_000.hdf

copy of the stack of 2D projections with Eulerian angles found at trial #000 set in the header. In order to examine the structure, one has to do the 3D reconstructions sxrecons3d_n.py outdir/structure_000.hdf myvol.hdf

outdir/structure.hdf

for multiple trials, it is a copy of the stack of 2D projections with Eulerian angles found at the best trial set in the header.

Description

Method

As described in the reference paper below, with modification of a full 2D Voronoi diagram on a sphere.

How to use the function sxfind_struct:

If the input stack contain already angles information (phi, theta, psi) for each projection, copy your file to a backup, because the program will overwrite the parameter xform.proj. In order to select the accuracy of the angle search delta, some examples below show the number of angles used in the list according the delta value:

• delta=20 list of 91 angles evenly distributed
• delta=10 list of 387 angles evenly distributed

To get a quick view of the structure found, use this function (sxrecons3d_n) to reconstruct the volume:

• sxrecons3d_n.py prj_stack.hdf myvol.hdf

The structure found depends on the random initialization of the program, controled by the option value random_seed. To improve the structure run many times the algorithm with different seed values. Select the best structure according the criterion Discrepancy wrote in the logfile. A lower criterion mean a better structure.

Reference

Penczek, P., Zhu, J. and Frank, J.: A common-lines based method for determining orientations for N>3 particle projections simultaneously. Ultramicroscopy 63:205-218, 1996.

Author / Maintainer

Julien Bert

Keywords

category 1

APPLICATIONS

Files

applications.py

See also

Maturity

stable

works for most people, has been tested; test cases/examples available.

Bugs