Emicofv
The program contains several options for searching for the view using
the common lines residual with varying weightings of the lines.
- Brute force scan of whole asymmetric unit by 1
degree steps in theta,phi,omega
- Limitted search in theta, phi and omega
These two searches use a chi-sqruare
calculation to estimate the quality of
the fit and compare different
orientations. This is done in RSYM and
produces several residuals.
- -A straight phase resid
- -a symmetry phase residual for views
near an axis
- -a probability of obtaining such
a phase residual estimated from the
distribution of residuals
- -a combination of the
probability and the symmetry resid
- -a vector residual - ie err A**2+errB**2
- -probability of vector resid
- -symmetry vector resid for views
near axes.
- -combination of these two vector
measures
RSYM - Details of probability calculation
This routine has been extensively modified by S. Fuller in june
1985 to overcome the problem of degeracies of common
lines in projection which lead to improper relative
weighting of the residual for different orientations.
The program now calculates the chi square for each
orientation and the probability of that chi square
based on an estimate of the variance obtained from
the matrix of residuals for a fixed theta and phi and
varying omega and the number of degrees of freedom
which is determined from the number of actual points
for which residuals are evalutated and the number of
samples in the complete line.
The chisqr is generated by evaluating the average value
of the residual on each common line and taking the
deviation (in the -'ve direction - so that a one
tailed test of chisquared is used ) as a measure of
the significance of the particular fit to the other
fits in the unit cell. The probability is calculated
on the basis of this chisqr and the number of occupied
common lines which went into its calculation and then
multiplied by the value of the residual hence low
residuals which are highly probable will be virtually
unchanged while those which are very improbable but
have the same value will be reduced.
The degeneracy of the common lines is compensated in several
ways.
- A flag which identifies truely degenerate lines
is set (IFLAG(NVIEW)) by ico_comlin_a.
- The distance between the lines is also
calculated or rather the radius for which
the common lines will be 1 transform unit
seperate (RAD(NVIEW)) by ico_comlin_a and the
search is made using only those points for
which this distance is greater than a
minimum (RADMIN)
- Modified to accept arrays of the phi and theta values for
the search so that the routine can be used for limitted
or brute force searches through calls to RESID2 and
- RESID3 respectively.
Made the value of IFLAG which is taken as indicating
a good pair of common lines a reported parameter.
(MIN_IFLAG)
- Since portions of the transform are not populated a trap
has been included to ignore residuals which do not have
valid NFREE or NSFREE values. They are included in the
calculation of sigma sqrd
A set of routines which calculate the phase and vector
residuals for fits to the symmetry near axes has
been added. The 2, 5 or 3 fold symmetry is computed
for that portion of the plane which lies within
1/2*radmin of the symmetry axis plane. This is
added to the prob chisqr corrected phase resid in
proportion to the number of points used in making
the determination of each.
SYMPLANE, DELZ, SYM_TWO_FOLD, SYM_THREE_FOLD,
and SYM_FIVE_FOLD
The phase residuals for centrosymmetry are doubled so that
they have the same average value as those for other phase
comparisons.
The effect of the probability calculation is applied to
the plane and the common lines residuals. The effect
is limitted so that a phase residual cannot be
decreased by more than a factor of 4.0 or increased
by more than a factor of 10. The tuning of the
probability for each new data set is done through
the parameters RPROB_FACT and SPROB_FACT which
should be adjusted so that the average probability
times _FACT is about 1.0
The routine calculates:
- a pure phase difference - ie the original
residual used in the Crowther program
(ARSMIN(ITHETA,IPHI))
- the phase chi square for which a probability
and a weighted chi square are also
calculated
- (RESMIN(ITHETA,IPHI),RPROB(ITHETA,IPHI),RESMIN/RPROB)
the vector (a*a+b*b) chi square for which a probability
and a weighted chi square are also
calculated
(SYMMIN(ITHETA,IPHI),SPROB(ITHETA,IPHI),SYMMIN/SPROB)
all seven of these are printed as they are arrayed in
the asymmmtric unit by RESREAD
IPRINT - Output control -
- if=3 unit 1 output
- if=2 then all unit 6 output
- if=1 then only normalized output
maps
- > MINR,MAXR,NSAMPL,FMIN,IPRINT (3I,F,I)
- MINR,MAXR (2i): INNER AND OUTER RADII (IN TF STEPS) OF BAND *
OF DATA THOUGHT TO BE CORRELATED ICOSAHEDRALLY *
- NSAMPL (i): # OF RADIAL SAMPLE STEPS TO BE TAKEN WITHIN THE *
BAND *
- FMIN (f): FRACTIONAL MINIMUM AMPLITUDE (RELATIVE TO MEAN) OF *
DATA USED IN THE REFINEMENT. DATA POINTS WITH SMALLER AMP- *
LITUDES ARE IGNORED. *
- FOR FMIN >0 THE AVERAGE VALUE OF F IN EACH RADIAL ZONE IS *
CALCULATED AND THESE VALUES ARE FIT , VIA A WILSON PLOT *
TO A STRAIGHT LINE TO ESTABLISH THE EXPECTED AVERAGE *
VALUE OF F WHICH IS THEN USED AS A CUTOFF. *
- FOR FMIN <0 THE AVERAGE VALUE OF F FOR THE ENTIRE AREA *
REQUIRED AREA IS CALCULATED AND THIS CUTOFF IS APPLIED *
TO ALL THE F'S. *
- IPRINT - determines the amount of output
- 0 - minimal - just final result
- 1 - three residual maps of the possible 7
- 2 - all residual maps and printer map of tf
- 3 - 2 and puts residual values to a disk file
- > FFT FILE NAME (A)
- > ORIGIN OK? (I)
- > enter new origin ORIGX,ORIGY(2F)
- > BRUTE FORCE SEARCH (I)
- to bypass or 0
- to do full 1 deg search of asym unit 1
- to do coarser search >1 (so 4 gives a scan in 4 steps
between 69deg and 90 deg
- > HEADER FOR TMP FILE (A)
this is the file generated in the 3 output option
but it is also put atop the asymmetric unit output
of residuals.
- > RADMIN (F)
Distance in recip space beyond which points are
independant usu 1/2*dia of particle
- >Limitted Search(I)
- to bypass or 0
- to perform 1
- > Theta: center,delta,number (2F,I)
- > Phi: center,delta,number (2F,I)
- > Radmin (F) - as above
*
- >OMEGA SEARCH (I)
- to bypass or 0
- to perform 1
- >Theta, PHI, Delta_omega (3F)
refines for a given theta, phi
- (at end asks for another omega search )
- >LIMITTED OMEGA SEARCH (I)
- to bypass or 0
- to perform 1
- >Theta, PHI, Delta_omega (3F)
- > Theta_start, Theta_end (2F)
- (at end asks for another limitted omega search, if 0 or
the program terminates )
com file
corresponding log file
log file for off center images
- Current Version 26-jul-95
-
- 26-jul-95 compiler alignment error in VAX version
- modified IPRINT from I2 to I4 to allow alignment of TITLE - sdf
- MODIFIED S. Fuller 16-juli-85 *
- Added option for removing the average value based on *
smoothed resolution dependence of F. (ZMIN) *
- Added printer output of the amplitueds and phases of the *
region used by the program (FOUROUT) *
- Upgraded to allow use of 512x512 transforms (ROT_FFT512) *
- Added calculation of phase residual from centrosymmetry *
as a function of resolution *
- Added option to allow coarse brute force search by *
responding to prompt with a value greater than 1 *
(RESID3) *
- Change output of orientation residual to generate an *
interpretable printer map (RESID3,RESREAD) *
- Calculate the distance in reciprocal space between pairs *
of common lines and use this to restrict search of *
orientations to radii where this is greater than a *
certain minimum value also flag those pairs of common *
lines which are either degenerate or are Friedel mates *
(RESID3,ico_comlin_a) *
- Changed calculation of orientation residual to a *
chi square residual and modify the residual by *
normalizing by the probability of chi square *
(RESID3,PCHISQ,MSDEV) *
- Change the calculation of the limitted orientation *
search to be consistent with the chisquare approach in *
the brute force asymmmetric unit search (RESID2) *
- Insert COMMON /COMPR/ to control the massive amounts of *
output which this set of programs produces *
- IPRINT: *
- <0 produces only best orientation and *
residual *
- =1 just produces the normalized residual *
maps and the mimimal set of outputs *
describing the run *
- =2 produces all the unit 6 output *
the printer map of all residuals and *
the descriptions of phase residuals and *
average amplitudes and the printer map *
of the transform *
- =3 produces all of above and outputs summary *
information to unit 1 *
*
*
*
- Remember: There is no such thing as a debugged program *
- There is only unrealistic test data. *
- further modifications to accomodate the MRC format for transforms
and to clean up the output handling (a bit.)
sdf 2--jan-87
*
- modified - sdf 15-feb 87 *
- modified in aug 87 to include automatic scaling of the
pobabilities and to output top values for each residual
- modified in nov 87 to put in more efficient prob(chisqr) calc
which does not require real*16 arithmetic
modified - 3-nov-87 - sdf
- modified in mar88 to correct bound bug
- needed to change the cosd(_) to cos(0.017* ) to avoid an
AXP optimization error
sdf 11-aug-94
Author: Stephen Fuller
=======================>
Return to Top Virus Page