Outline for Real Space averaging of Helical Tubes I. Average edited data set in Fourier space HAVGREFN: to refine fitting parameters of individual images HLXAVG: to create average data set (id.avg and ref.avg) LLPRX2: examine averaged data set and select only strong layer lines with good 2-fold residuals for determination of fitting parameters. HLXEDT: select strong layer lines and limit extent idEDTRES.avg notes: use idred.2fd to start that has been adjusted to correct 2-fold origin to ultimately be used for map averaging (of difference calculation) Two-fold for averaging data sets is between dimer ribbons Two-fold for difference map is on vanadate peak II. Corrections to averaged data set CTF (HDIVCTF) -> idEDTRES.cor and refEDTRES.cor MTF (MTFCORLL) -> idEDTRESm.cor and refEDTRESm.cor Limit resolution for initial fits (HLMTRES) idmEDTRES.cor and refmEDTRES.cor III. Determine radial scaling factor against reference data set (2 ways possible) A. Using MRDD HLX2FLD: calculate two-fold enforced file (idEDTRESm.2fd and refEDTRESm.2fd) HLXFB: calculate little g file with 1A spacing (idEDTRESm.lg1 and refEDTRESm.lg1) LGLST: calculate MRDD for both idEDTRESm.lg1 and refEDTRESm.lg1 GNUPLOT lglst.gnu: plot MRDD LGCOR: edit lglst.dat to use as control file -> RSCAL, RSHIFT notes: adjust scale for lglst so that MRDD have roughly the same scale use cutoff amplitude for LGCOR so that only features of MRDD are used for correlation, not just basic square-wave shape B. Using COMPMOL make HPZ sections (see below) using either RSCAL=1 or LGCOR values for RSCAL use RHOFIT control file to run COMPMOL to get RSCAL value notes: LGCOR doesn’t seem to work very reliably if the MRDD has different shape between the two data sets (i.e., lipid bilayer peak spacing smaller and distance to cytoplasmic head peak bigger as was our case). So we ultimately decided COMPMOL gave a more reliable result. IV. Calculate HX sections HLXAVG: (fitid.cnt) to apply RSCAL to data set (idmEDTRESr.avg) HLXFB: calculate little g file with 3A spacing from non-two-fold enforced file (idmr.lg3 and refm.lg3) HXSEC: calcluate HX sections (idEDTRESrm.xn and refEDTRESm.xn) odd number of pixels in all dimensions choose limits for map to include only one molecule (consistent with ref) use RPT/RSCAL for repeat distance notes: consider using 1A spacings even for alignment purposes to avoid confusion and inaccuracies when switching over to unedited data. V. Define mask for scaling Mask unit cell for scaling HCELLBOX (requires img.id and ref.id) creates cnt file for boxmask BOXMSK: apply mask to X sections (idmEDTRESr.msk) Close molecular mask IMGCHR: convert masked X sections to character format (idEDTRESm.chr) VOLUME: calculate density cutoff for 100% volume recovery IMGDENS: apply mask with this density cutoff (idEDTRESmr.msk -> idEDTRESmr.dns) notes: close molecular mask, although theoretically rigorous, may not work well. So unit cell is ok for scaling VI. Scale two data sets IMGCORAB: calculate linear regression for densities (idEDTRESmr.dns vs. refEDTRESm.dns) or (idEDTRESmr.msk vs. refEDTRESm.msk) GNUPLOT ab.gnu: check linear regression PRODADD: apply density scale to data set (idEDTRESmrs.dns) notes: Best scale factor should consider only densities from the molecules themselves and not background densities. However, IMGCORAB doesn’t always give a good correlation with such a tight mask. Should examine gnuplot plot of least squares regression to see if fit is acceptable. VII. Align bottom molecule with reference data set XDISPMSK: define rough mask around bottom molecule of dimer, both test and reference images (idEDTbottom.bmk and refEDTRES.bmk) BOXMSK: apply mask to bottom molecule (idEDTRESbottom.msk) RHOFIT: crosscorrelation to determine relative orientation (idEDTRESbottom.msk vs refEDTRES.msk) COMPOL: similar to RHOFIT except RSCAL also considered, can be used to determine radial scale when LGCOR seems unreliable (i.e., big enough difference in shape of MRDD) SKEW: apply fitting parameters determined by RHOFIT to scaled data set (idskewEDTRESbot.msk) notes: 1. density scale probably not so important for this because correlation coeff is independent of scale. However, if negative densities are present in one case and not another, could be trouble. So best to be about the right density scale 2. check skewed map against reference map to make sure you are happy with relative orientations and radial scale and density scale factors. Also can use HISTO-0 to compare density distributions and unbiased mean density (ignores zero density) for two maps. 3. SKEW not actually necessary, just to check whether fitting was successful 4. rough mask better than 100% volume mask because edge of molecule used during correlation, which is a powerful constraint for finding the relative position (zeros in image ignores during correlation) VIII. Align top molecule with reference data set BOXMSK2FLD: generate 2-fld related mask (idEDTRESTOP.bmk) BOXMSK: apply mask to bottom molecule (idEDTREStop.msk) RHOFIT: crosscorrelation to determine relative orientation (idEDTREStop.msk vs refEDTRES.msk) SKEW: apply fitting parameters determined by RHOFIT to scaled data set (idskewEDTREStop.msk) VIII. Prepare unedited data files for averaging (use parameters for edited data) HLXAVG -> id.avg and ref.avg HDIVCTF -> id.cor and ref.cor MTFCORLL -> idm.cor and refm.cor radial scale HLXAVG fitid.avg -> idr.avg HLXFB -> idmr.lg1 and ref1.lg1 (1A intervals) HXSEC -> idmr.hxn and refm.hxn (no two-fold enforced, 1 A intervals) PRODADD (apply scale ) -> idmrs.hxn SKEW (adjust orientation of top and bottom independently) -> idlowerskew.xn and idupperskew.xn notes: very important that the center of these 1 A X-section maps be the same as the 3A maps. Currently there are bugs in HXSEC that make this difficult, so be very careful. May need to use LABEL and ALTERNXYZST to fix HXSEC result: LABEL screws up the start section number which must be restored. Use MRCHD to check header info at all stages: idupperskew.xnc and idlowerskew.xnc. IX. Mask and Sum data sets (top and bottom handled independently) HCELLTWO makes control files: lower.cnt and upper.cnt for masking upper and lower half of unit cell BOXMSK (using lower.out and upper.out) -> idupperskew.xnc2 and idlowerskew.xnc2 AVGIMTWO: run twice with avglower.cnt and avgupper.cnt notes: AVGIMTWO calculates sum of all pixels. If pixels in both maps nonzero, outputs average, if pixels in one map is zero, then outputs sum. After applying twice to both halves, result if independent average of upper half and lower half with reference. Note that AVGIMROT outputs zero density if density from either map is zero (i.e., uses one of maps as a mask for the other). Note that program could be changed to just take simple sum (no average and no testing for zeros) and get the same result * 2. --------- --------- --------- --------- --------- |upper /| | | |upper /| |0 0 0 /| |upper /| |id / | | | |avg / | | 0 0 / | |avg / | | / | | | | / | |0 0 / | | / | | / 0 | + | ref | = | / | + | 0 / | = | / | | / 0 0| | | | / | |0 / | | / | | / 0 0 | | | | / ref | | / id | | / avg | |/ 0 0 0| | | |/ | |/ lower| |/lower | --------- --------- --------- --------- --------- X. Back to Fourier Space to edit layer lines HX2LG: hx2lg.cnt to calculate little g's from averaged x-section -> id+ref.lg1 Make big G's (layer line data) HLXTBL to create *.id file as template which must have full extent of layer lines. So specify refid.avg file (unaveraged, full length layer line data from image that defines the indexing of reference data set) and produce ref.edt file which is used by next program (watch out for blank line just after title) HLXFBR: to make big G file (layer line data) XI. Edit layer line data to produce map HLMTRES: maybe to put hard limit on resolution LLPRX2: to remove bad data from layer lines HLX2FLD: to enforce 2-fold symmetry HLXFB: to create 1A little g file avg.lg1 HXSEC, or HFBSEC or HCLND to create map Utility programs: MRCHD: print out header info on MRC image file LGHD: print out header info on little g file LLDATAHD: print out header info on layer line data SAVGLG: prepare plot for various aspects of averaged layer line data Real-space averaging 06/27/99 1