(last update Jun 20, 2002)

EM3DR is normally run after and in combination with EMPFT or EMICOGRAD.

In *THEORY*, EM3DR has no restrictions as to number of particles combined or the resolution limit. Of course, the memory capacity of the computer and disk space available WILL pose ultimate limits if you go overboard.

Latest changes (Jun 2002): Option for different CTF correction filters.

1. MODE, SYM, BIN, IPRINT, BG_OUT, LG_OUT, MP_OUT (7I)

2. RES_HI,RES_OUT,UNITS,MAP_MAGFAC,MAP_DIM,ICO_SPREAD (2F,I,F,I,F)

4. BG data input/output filename (A) (record ignored if MODE = 0)

5. 3D MAP output filename (A; DEFAULT = ICOS2F.PIF)

6. Header for 3D MAP (18A4)

7. PARAMETERS filename(s) (A)

1. MODE, SYM, BIN, IPRINT, BG_OUT, LG_OUT, MP_OUT (7I)

----------------------------------------------

Input parameters control the operation of EM3DR as follows:

MODE = 0 Entire program is run to compute a 3D map from images

= 1 Stop after calculating BG'S (stored in BG.DAT)

= 2 Start with existing BG.DAT data and compute 3DR

SYM = 1 if no symmetry

2-31 if n-fold about z-axis

532 if icosahedral

The following dihedral point group symmetries are allowed:

222,32,422,52,622,72,822,92

BIN = 1 Binfactor of 1 (DEFAULT; i.e. no data compression)

= 2 Binfactor of 2 (compresses 3D data eightfold and 2D

data fourfold) The current limit on BIN is 2.

NOTE: BIN=2 is mainly designed to be used in conjunction with

EMPFT when it is used with BIN=2. When BIN is set = 2,

then MAP_MAGFAC defaults to 0.5, regardless of any other

value you may enter for MAP_MAGFAC.

IPRINT = 0 gives minimal output to terminal or *.LOG file

= 1 gives extra output

= 2 generates lots of output!!!

BG_OUT = 1 BG data saved in binary file 'BG.DAT'

LG_OUT = 1 LG data saved in binary file 'LG.DAT'

MP_OUT = 0 - output whole 3D map in standard 2-fold view

= 1 - use only odd set of images to compute MAP

= 2 - use only even set of images to compute MAP

When MP_OUT = 1 or = 2 then only EVERY OTHER image in the total input set of images is used for the calculation of the 3D MAP. Note that "odd" or "even" does NOT mean that images are selected according to particle ID, but instead just means that the particle images are selected according to the SEQUENCE in which they are read from the PARAM file(s). For example, if your FIRST PARAM file only lists particles with even-numbered IDs (e.g. 22, 34, 36, 58, 100, 268, 982, 1046, etc.) and MP_OUT is set = 1, then only the images with ID #s 22, 36, 100, 982, etc. will be used. Also, if in this last example, the PARAM file was instead the 2^nd PARAM file listed in the EM3DR batch job AND the first PARAM file specified an ODD number of images, then only images with ID #s 34, 58, 268, 1046, etc. would be selected from this PARAM file. This strategy assures that, no matter how many PARAM files there are and no matter what the particle IDs are, the number of images selected will always be equal to or within one of being half of the total number of images listed in the PARAM files.

When MP_OUT = 5, only a top (near) quadrant of the 3D density map in the five-fold view is computed and stored in a disk file. Choosing this option may be very useful for producing preliminary maps where you just need a quick look at a partial map to see if it is correct.

Future enhancements to EM3DR may include options to write out even smaller portions of the standard 3D map (e.g. just a quadrant) which will generate greater disk storage savings and perhaps enhancements in other routines.

2. RES_HI, RES_OUT, UNITS, MAP_MAGFAC, MAP_DIM, ICO_SPREAD

(2F,I,F,I,F)

-------------------------------------------------------

RES_HI specifies the highest possible resolution to which the final 3D map can be computed. RES_HI (and RES_OUT) are expressed in units defined by the user (see description of UNITS, below). RES_HI can't be set to a value any SMALLER than twice the size of the IMAGE pixels (the so-called Nyquist limit imposed because the Fourier transform of digitized data only extends to a spatial frequency of 1.0/(2*pixel size). If you believe your IMAGE data are good enough to achieve higher resolution, you need to rescan the IMAGE at smaller pixel resolution. The typical rule of thumb is to digitize your IMAGE data at a pixel resolution that is 3 or 4 times the expected resolution, in which case only information in the Fourier transform out to a radius (IRAD) of FFT_DIM/3 or FFT_DIM/4 is used. If, for example, you expect your IMAGE RESOLUTION to be about 2.0 nm, then you would normally digitize at 0.7 nm intervals (or smaller).

RES_OUT is the resolution to which the final 3D map is computed. RES_OUT must be equal or larger than RES_HI. Normally RES_OUT is set equal to RES_HI (the DEFAULT), but sometimes it is useful to calculate several reconstructed density maps at different resolutions. This can be done efficiently by calculating and saving the BG data (MODE = 1) with RES_HI set to the maximum resolution desired and then running EM3DR (MODE = 2) to reuse the already calculated BG data, but setting RES_OUT to different values (larger than RES_HI) for each new map you calculate.

UNITS specifies the units assigned to the variables RES_HI and RES_OUT and also to the PIXSIZ variable obtained from each of the PARAM files (see description of PARAM files below: input line #7). The user may define UNITS according to the following table, but REMEMBER - make sure you are CONSISTENT and use the same UNITS to specify ALL these variables.

0 = pixels (i.e. dimensionless)

2 = nanometers (nm)

MAP_MAGFAC is used to adjust the size of pixels in the final 3D map and also (if MAP_DIM is left = 0) will automatically resize the dimensions of the 3D map. The DEFAULT (1.0) will make a 3D map with pixels identical to the SMALLEST pixels in the input IMAGE data and the map will have NCOL**3, dimensions, where NCOL is the LARGEST linear dimension of the input IMAGE data. Set MAP_MAGFAC > 1.0 if you wish to produce a larger than normal reconstruction, one that is sampled with smaller size MAP pixels to achieve much nicer 3D rendering effects for making high quality figures for slides or publication. For example, if MAP_MAGFAC = 2.0, the final 3D map will have dimensions (2*NCOL+1). Under these conditions EM3DR will take considerably longer than normal to finish and may 'consume' very large amounts of disk space to store the finely sampled output map. Typical, final reconstructions used to generate nice surface-shaded renderings (e.g. with ROBEM) are computed with MAP_MAGFAC = 2.0. If you wish to get a map quickly and don't mind coarse MAP pixel resolution, set MAP_MAGFAC < 1.0 and the 3D map will be correspondingly smaller and take up much less disk space (and be rendered much faster, etc.).

MAP_DIM can be used to specifically fix the dimensions of the 3D map and thereby override the effects of any change in map dimensions brought about by using a value for MAP_MAGFAC other than 1.0. Though not likely to be used too often, MAP_DIM adds flexibility to how you create output 3D map data. Setting this parameter can be useful, for example if you are trying to produce a 3D model of the right dimensions for use in EMPFT, which currently requires that the NCOL and NROW dimensions of the 3D model EXACTLY match the NCOL and NROW dimensions of the input IMAGE data. MAP_DIM may also be used to keep the dimensions of the 3D output map constant, while at the same time allowing the pixel size of the 3D map data to be magnified or demagnified according to the value of MAP_MAGFAC. For example, you may choose to calculate a 3D map with, say MAP_MAGFAC = 0.75, but with MAP_DIM set equal to the size the 3D map WOULD have been IF MAG_MAGFAC were left = 1.0. This would give a 3D map in which the reconstructed particle would appear shrunk by 25% but still inside a 3D box of dimensions equal to the dimensions of the original input IMAGE data.

WARNING: If you decide to use MAP_DIM, beware of possible "side effects". For example if you set it too small depending on what MAP_MAGFAC is set to, the map may be cut off beyond a radius that lies within the particle! Sooooooooo, unless you really need to use this option, leave MAG_DIM = 0 and let the program size the 3D output map according to the value of MAP_MAGFAC and the input IMAGE dimensions.

ICO_SPREAD is a seldom used variable that sets the condition on the mean inverse eigenvalue. This is usually (99.99% of the time) set equal to 1.0 (The DEFAULT). Set to a lower value for a less stringent test.

-------------------------------

These parameters are used to affect CTF corrections of the images. The following is a simple description of the nature of the use of these variables. More information will be supplied as we become more adept at using CTF corrections with real data.

CTFMODE is a switch that signifies the type of CTF correction to make.

CTFMODE Action taken

------- ------------------------------------------------------

0 No CTF correction made to the FFT data.

1 Full CTF correction (flip phases and modify amplitudes)

2 Only flip phases in the FFTs

3 Only modify amplitudes in the FFTs

4 Multiply image by CTF (WARNING) blah blah

FILTER options and formulas

FILT_FAC is the FACTOR TO MAKE SURE NOISE IS NOT MAGNIFIED TOO MUCH NEAR THE NODES OF THE CTF. THE FORMULA USED depends on the value of FILTER:

FILTER FORMULA

------------------------------------------------------------------------

1 FILTER 4 (before first peak)+ FILTER 2 (after first peak)

2 FFTnew = (FFTold*ABS(CTF)*TFAC)/(CTF**2 + FILT_FAC*(1-ABS(CTF))

3 FFTnew = (FFTold*ABS(CTF)*TFAC)/(CTF**2 + FILT_FAC)

4 FFTnew = (FFTold*TFAC)/(ABS(CTF) + FILT_FAC*(1-ABS(CTF))

5 FFTnew = (FFTold*TFAC)/(ABS(CTF) + FILT_FAC)

4. File containing BG data (A; DEFAULT = BG.DAT)

-----------------------

Enter the name of the permanent file in which the big G data are stored (MODE=2). This input line is read but ignored if MODE.LE.1.

5. 3D MAP output filename (A; DEFAULT = ICOS2F.PIF)

----------------------

Specify the name of the PIF format file to store the 3D map. This input line is read but ignored if MODE=1.

6. Header for 3D MAP (80A1)

-----------------

The header is text you can use to identify the 3D map data. This input line is read but ignored if MODE=1.

7. Input PARAMETER filename(s) (A FORMAT)

---------------------------

ID is the image number--the storage position of the IMAGE data for byte-packed format data.

THETA,PHI,OMEGA are the current values of the three orientation angles (in degrees) for each particle image (obtained from previous runs of EMPFT or EMICOGRAD or EMICOFV).

FFT_ORIGX,FFT_ORIGY are the pixel coordinates of the particle center (the point 0.0,0.0 corresponding to the lower left corner of the boxed particle image). These values are estimated or refined by EMPFT.

MAG_FAC is a radial scale factor computed in the EMPFT program. It specifies the size of each particle image relative to some standard, such as a 3D reconstruction model. Hence, a MAG_FAC = 1.02 means that the particle image is 2% larger than the model. EM3DR uses this value to assure that all particle transforms are sampled so the data are all combined at a uniform magnification. With images of frozen-hydrated particles boxed from a single micrograph, MAG_FAC = 1.0 (DEFAULT) is normally assumed for all particles.

NOTE: Input line 7 and following are ignored if MODE=2.

Normal operation of EM3DR leads to a 3D that map contains the entire reconstructed particle viewed down a 2-fold axis in the standard orientation: for a particle with icosahedral (532) symmetry, this is where three mutually perpendicular two-fold particle axes are aligned with a Cartesian (XYZ) map coordinate system (NCOL columns in the X direction; NROW rows in the Y direction; NSEC sections in the Z direction). For particles with other symmetry, such as with Cn (cyclic) or Dn (dihedral) point group symmetry, the n-fold axis coincides with the Z-axis.

! COMMAND FILE FOR RUNNING EM3DR

!

! THE INPUT ARE AS FOLLOWS:

!

! 1. MODE, SYM, BIN, IPRINT, BG_OUT, LG_OUT, MP_OUT (7I)

! MODE = 0: run entire program

! = 1: stop after calculating bg's

! = 2: start with existing BG.DAT data

! SYM = 1 if no symmetry (like a ribosome)

! = 2-31 if n-fold cyclic symmetry (about z-axis)

! = 532 if icosahedral symmetry

! = 222 if 222 dihedral symmetry

! = 32 if 32 dihedral symmetry

! = 422 if 422 dihedral symmetry

! = 52 if 52 dihedral symmetry

! = 622 if 622 dihedral symmetry

! = 72 if 72 dihedral symmetry

! = 822 if 822 dihedral symmetry

! = 92 if 92 dihedral symmetry

! BIN = 1: no data compression (DEFAULT)

! = 2: compresses 3D data eightfold and 2D data

! fourfold). The current limit on BIN is 2.

! IPRINT = 1: extra output to terminal or *.LOG file

! = 2: even more output to terminal or *.LOG file

! BG_OUT = 1: BG.DAT saved (otherwise not saved)

! LG_OUT = 1: LG.DAT saved (otherwise not saved)

! MP_OUT = 0: output whole 3D map in standard 2F view

! = 1: use only odd-order images to compute MAP

! = 2: use only even-order images to compute MAP

! = 5: output top quadrant in 5-fold view

!

! 2. RES_HI, RES_OUT, UNITS, MAP_MAGFAC, MAP_DIM, ICO_SPREAD

! (2F,I,F,I,F)

! 3. CTFMODE, FILTER, TFAC, FILT_FAC (2I,11F)

! 4. BG data input/output filename (A) (ignored if MODE = 0)

! 5. 3D map output filename (A; DEFAULT = ICOS2F.MAP)

! 6. Header for 3D map (80A1)

! 7. Parameter filename(s) (A)

!

$ SET DEFAULT DEXTRO3:[TSB.V.HSV.TEST]

$ RUN DEXTRO3:[TSB.EXE]EM3DR.EXE

1, 532, 1, 0, 0, 0, 0

35., 35., 1, 1.0, 0, 1.0

0, 1, 0.0, 0.1

HSV_BG.DAT

HSV.PIFMAP

Equine herpes intermediates #1856 n=60 3DR 99**3 (35A resolution)

HSV.DAT

$ WAIT

$ EXIT

Several parameters are listed at the terminal (or in the *.LOG file):

FFT_DIM

------------- TPU

BOX_DIAM*2.0

For ICOSAHEDRAL particles, EM3DR first computes one half of the density map, oriented with a five-fold axis coincident with the Z axis and one, perpendicular 2-fold axis coincident with the Y direction, from the little g's. Little g's are only computed for Bessel orders that are multiples of 5. The density map is calculated a section at a time in unique 2p/5 (= 72°) sectors. Each sector is subdivided in cylindrical coordinates (RHO(r,f)) since the little g's are computed in polar form. The polar section is interpolated to produce a Cartesian section (MAP(x,y)).

* |- BGSOLVE ------|- CGESV (Math library routine)

* |- BGOUT ! |- CGEEV (Math library routine)

* |- BGSOLVE2 -|- ICO_HOW ------|- ICO_EIGVAL !

* |- PIF_WRITE_SECI2 - PIF_WRITE_MAP_SECTION**

|- APPLY_3F** |

|- PIF_WRITE_SECI2 - PIF_WRITE_MAP_SECTION**

| |- PIF_READ_WMAPI2 - PIF_READ_MAPI2**

| |- PIF_WRITE_SECI2 - PIF_WRITE_MAP_SECTION**