Example

Session


The following is a (nearly) verbatim session with Elves that was used to solve a selenomethionine derivative of dnaG:

Keck JL, Roche DD, Lynch AS, Berger JM. "Structure of the RNA polymerase domain of E. coli primase". Science. 2000 Mar 31;287(5462):2482-6.

It is repoduced here with the kind permission of Jim Keck and James Berger.

Text produced by Elves is in white, and text typed in by the user is indicated in green. Interspersed commentary from me (not part of the actual session), are in blue. Text in <braces> is something the user typed that doesn't get displayed (like the Enter key).


This session begins with an email delivery of the Elves program (to myself)

Last login: Tue Aug 31 10:29:14 1999 from machine.berkeley.edu
You have mail.
unix 1% cd /laue/jamesh/showoff/Elves/
unix 2% mail
Mail version SGI.950426.  Type ? for help.
"/var/mail/jamesh": 1 message 1 new
>N  1 jamesh@ucxray.berkeley.edu     Tue Aug 31 12:09  
& s Elves
"Elves" [New file] 44328/1358509
& d
& q

This command removes the email header from the Elves script

unix 3% awk '/#! /,""' Elves >! elves
unix 4% mv elves Elves 
unix 5% chmod a+x Elves 

Here is where I actually run Elves

unix 6% ./Elves data in /data/semet/ is from the protein in ~/orf.seq
Elves v 0.9        Because you have better things to do.(TM)  James Holton 8-31-99

Cataloging frames in /data/semet/ .......
Getting protein sequence from  /usr/users/alber/jamesh/orf.seq
Hang on ...

ready to set up processing directories for:

single crystal of:
36482.5 D chain: 
YMRDGAGNLPQSFARYEKQGGRSWIYHQQNLLVAQNNQLFMGTGSFLVRSIRRQLTGTAVFDIKPLTPHPAAVDLEQEQL
EQGDDEPMRRVESDLRSLIKLIGNYYLDRRYPDRKYFFTGLKNNGQYRLVRVEGDRVQDVEFPMPHAQGNGIGATNVFRG
ATNPIACVDYSRDVHFAQFYADQRALEIRTYTLRADTPILDNLGPGYSTRWGLLCDMFAEDTAQDLGRMTVAAALLVNGG
LELTGLLEVFLTQGMALYDRSDSAQSSARMQFAMGLEPEELVLNKARKPPFQPRDISQRDLRLGEGLKETSIILARNLPL
KPM

Total: 36482.5 D

directory Energy  First_Frame               total   phi0    osc distance  x_beam  y_beam 2theta 
wave1     13000.0 /data/semet/hr_1_001.img    180  180.0    1.0   240.00   93.14   94.20    0.0 
wave2     12658.0 /data/semet/pk1_1_001.img   180  180.0    1.0   240.00   93.14   94.20    0.0 
wave3     12663.0 /data/semet/in1_1_001.img   180  180.0    1.0   240.00   93.14   94.20    0.0 

in unknown space group.
with 10 Selenium sites / 36.5 kD
Expect: Riso= 5.2% Rano= 0.1% Rdisp= 0.5% at 12660.0 eV

Everything look okay? [Yes]
-> no
What's wrong? [nothing]?
-> the beam center is really 93.1 95.0 and there are only 8 sites
Hang on ...
Is this beam center from Mosflm, adxv, R-axis(horiz, vert), or denzo? [mosflm]
-> <Enter>

ready to set up processing directories for:

single crystal of:
36482.5 D chain: 
YMRDGAGNLPQSFARYEKQGGRSWIYHQQNLLVAQNNQLFMGTGSFLVRSIRRQLTGTAVFDIKPLTPHPAAVDLEQEQL
EQGDDEPMRRVESDLRSLIKLIGNYYLDRRYPDRKYFFTGLKNNGQYRLVRVEGDRVQDVEFPMPHAQGNGIGATNVFRG
ATNPIACVDYSRDVHFAQFYADQRALEIRTYTLRADTPILDNLGPGYSTRWGLLCDMFAEDTAQDLGRMTVAAALLVNGG
LELTGLLEVFLTQGMALYDRSDSAQSSARMQFAMGLEPEELVLNKARKPPFQPRDISQRDLRLGEGLKETSIILARNLPL
KPM

Total: 36482.5 D

directory Energy  First_Frame               total   phi0    osc distance  x_beam  y_beam 2theta 
wave1     13000.0 /data/semet/hr_1_001.img    180  180.0    1.0   240.00   93.10   95.00    0.0 
wave2     12658.0 /data/semet/pk1_1_001.img   180  180.0    1.0   240.00   93.10   95.00    0.0 
wave3     12663.0 /data/semet/in1_1_001.img   180  180.0    1.0   240.00   93.10   95.00    0.0 

in unknown space group.
with 8 Selenium sites / 36.5 kD
Expect: Riso= 4.1% Rano= 0.1% Rdisp= 0.4% at 12660.0 eV

Everything look okay? [Yes]
-> I want to edit this table
Do you want to exit and edit wedges.txt [Yes]
-> yup
Please edit wedges.txt with your favorite text editor.
(Remember to keep the wedge table nice and vertical)

When you are done, type:
./Elves wedges.txt
unix 15% vi wedges.txt 

You don't have to use vi. At this point, I just wanted to give the wavelength directories more sensible names. All you have to do is make sure the wedges.txt file still looks "nice" (with vertically-aligned columns) when you save it. Once the wedges.txt file describes the way you collected your data, you can launch Elves again:

unix 19% ./Elves wedges.txt
Elves v 0.9        Because you have better things to do.(TM)  James Holton 8-27-99

Getting image file layout from wedges.txt
Getting protein sequence from  wedges.txt
Hang on ...

ready to set up processing directories for:

single crystal of:
36482.5 D chain: 
YMRDGAGNLPQSFARYEKQGGRSWIYHQQNLLVAQNNQLFMGTGSFLVRSIRRQLTGTAVFDIKPLTPHPAAVDLEQEQL
EQGDDEPMRRVESDLRSLIKLIGNYYLDRRYPDRKYFFTGLKNNGQYRLVRVEGDRVQDVEFPMPHAQGNGIGATNVFRG
ATNPIACVDYSRDVHFAQFYADQRALEIRTYTLRADTPILDNLGPGYSTRWGLLCDMFAEDTAQDLGRMTVAAALLVNGG
LELTGLLEVFLTQGMALYDRSDSAQSSARMQFAMGLEPEELVLNKARKPPFQPRDISQRDLRLGEGLKETSIILARNLPL
KPM

Total: 36482.5 D

directory Energy  First_Frame               total   phi0    osc distance  x_beam  y_beam 2theta 
hr        13000.0 /data/semet/hr_1_001.img    180  180.0    1.0   240.00   93.10   95.00    0.0 
inf       12658.0 /data/semet/in1_1_001.img   180  180.0    1.0   240.00   93.10   95.00    0.0 
peak      12663.0 /data/semet/pk1_1_001.img   180  180.0    1.0   240.00   93.10   95.00    0.0 

in unknown space group.
with 8 Selenium sites / 36.5 kD
Expect: Riso= 4.1% Rano= 0.1% Rdisp= 0.4% at 12660.0 eV

Everything look okay? [Yes]
-> <Enter>
Good.
 Don't go away ...
/laue/jamesh/showoff/Elves/hr/wedge1 is for /data/semet/hr_1_001.img
/laue/jamesh/showoff/Elves/inf/wedge1 is for /data/semet/in1_1_001.img
/laue/jamesh/showoff/Elves/peak/wedge1 is for /data/semet/pk1_1_001.img

deploying elves into /laue/jamesh/showoff/Elves/Elfsheim ......

./Elfsheim/Wedger will index and process one wedge of data.
./Elfsheim/Scaler will localscale and merge multiwedge data.
./Elfsheim/Phaser will refine sites and calculate phases.
./Elfsheim/Processer will process all your data.

Shall we index your first wedge? [Yes]
-> y

moving us into indexing directory...
cd ./index

Here we launch Wedger Elves. The purpose of this run is not so much to completely process this first wedge, but to refine the crystal and camera parameters to guide the processing of all wedges. It is usually a good idea to process this first wedge "manually" (with Wedger Elves) in order to set up things like the beamstop shadow, enter any project-wide mosflm keywords, and to be confident that spots are actually being predicted and the refinement is stable.

Wedger: v1.0b      Because you have better things to do.(TM)   James Holton 8-30-99

found: ./start


Scanning /data/semet/hr_1_001.img ... 0.95372 Å 1-180 
Scanning ./start .....

Looking for ipmosflm ... found /programs/mosflm/ipmosflm
Checking frames (Ctrl-C to skip) ..................................................................................................................................

        About to autoindex: 
        /data/semet/hr_1_???.img

        using:       /programs/mosflm/ipmosflm v 6.01

        beam center: 93.10 95.00
        distance:    240.00 mm
        wavelength:  0.953725 Å
        cell:        unknown
        space group: unknown

Is this correct? [Yes]
creating autoindex.inp...

WARNING: non-interactive indexing cannot determine your space group exactly.
         (but, the elves can pick a close one, and reindex later.)

DPS indexing in MOSFLM Version 6.x is very cool,
but can only be done in the graphics window! 
Do you want to autoindex INTERactively (with graphics)? [Yes]
-> y
####################################################################################
WHAT TO DO NEXT:

1) Say "Yes" to the question below.

2) the big Mosflm graphics window should come up.  (if not, read the errors and
   make sure CCP4 is set up)

3) Click on "Autoindex"  follow the directions you see.

4) If nothing too scary happens Click on "Predict"

5) If you see little boxes around your spots, You WIN! (auto.mat is your new matrix)

6) Exit Mosflm, and come back here.

P.S.
You might also want to take this opportunity to define your backstop shadow
 using the "KEYWORDED INPUT" button and typing something like:
 "BACKSTOP CENTER 93.10 95.00 RADIUS 5"

####################################################################################
launch mosflm and do some autoindexing now? [Yes]
-> <Enter>

 ************ Version 6.01 for Image plate data 06/07/99  ***********
 A.G.W. Leslie, MRC Laboratory Of Molecular Biology, HILLS ROAD, CAMBRIDGE CB2 2QH, UK
 E-mail andrew@mrc-lmb.cam.ac.uk
 New auto-indexing using DPS due to Ingo Steller Robert Bolotovsky and Michael Rossmann
 (1998) J. Appl. Cryst. 30, 1036-1040
 Original auto-indexing using REFIX due to Wolfgang Kabsch (Kabsch,W. (1993),J.Appl.Cryst. 24,795-800.)
 X-windows interface using xdl_view due to John Campbell (Daresbury Laboratory, UK.)
 (Campbell,J.W. (1995) J. Appl. Cryst. 28, 236-242.

......


At this point, you would see the mosflm graphics window come up...


All you really have to do is hit the "Autoindex" button, but, I find, you will get a much better and more stable initial orientation if you use "Find spots" and "Read image" to pick spots on more than one image (four is usually plenty) before you hit "Autoindex". You should use images that are widely-separated in phi. Because it is a good idea to get some initial shots at phi=0 and phi=90 to verify your crystal centering anyway, you might as well use those two images for autoindexing. As long as you havn't moved the crystal, your indexing images don't have to be part of the wedge you are about to process.

Once you have done autoindexing, hit "Predict" and see if the spots on the image (and, if possible, a few other images) have prediction boxes over them. If they do, you can "EXIT" mosflm, and return to Elves. As long as you have predictions on your spots, and you aren't using a totally wrong lattice symmetry, Wedger Elves should be able to handle it from here.

*********** END OF PROCESSING *****************

There, now wasn't that fun? :)

What space group did you pick? [P212121]
-> <Enter>
Scanning auto.mat ... 38.1886 57.6016 147.314 90 90 90
Scanning hr_1_001.sav .....
Scanning logs/autoindex.log ....
Hang on ... 



Wedger elves will create a script called mosflm.com that will run /programs/mosflm/ipmosflm
         and process from /data/semet/hr_1_001.img
                       to /data/semet/hr_1_180.img
         using the orientation in auto.mat

        The Mosflm interactive graphics window will be launched.

        Data were collected from 180° to 360°, in 1° steps with the 
        ALS ADSC detector (gain = 0.3) at 240.000 mm from the crystal and
        with the direct beam hitting the detector at: 93.22 95.21

        X-rays were 0.95373 Å, with a spectral bandwidth of 1:10000, and 
        the beam was diverging by 0.100 x 0.020° with polarization 0.9

        Unit Cell is 38.1886 57.6016 147.314 90 90 90
        Spots will be measured out to 2.0 Å with space group P212121
        and a mosaic spread of 1.00°

        Refinement will be in 7 blocks of 4 frames. No integration.

        additional keywords:
        BACKSTOP CENTRE 94.00 88.00 RADIUS 10.00

           ^   ###  ####   #   ###    ##### #  # #  ###    ^
          /|\  #  # #     # #  #  #     #   #  # # #      /|\
           |   ###  ###  ##### #  #     #   #### #  ###    |
           |   # #  #    #   # #  #     #   #  # #     #   |
           |   #  # #### #   # ###      #   #  # # ####    |

      Make sure all the above numbers are correct.
           (Especially the beam center)

Wedger Elves will compute a collection strategy for you, if you like.

Do you want our opinion on your collection strategy? [Yes]
-> <Enter>
This wedge will be 100.0% complete to 2.0 A and contain 99.3% of anomalous data.
A wedge from  323.0 to 413.0 degrees would be 97.6% complete.
A wedge from  298.0 to 388.0 degrees would contain 63.4% of anomalous pairs.

At any point during Wedger Elves processing, you can type "graphics" or "window" and do the next mosflm run with the graphics window up.

Do you want to monitor this run in the graphics window? [No]
-> n
Shall we run mosflm.com now? [Yes]? 
-> <Enter>
Refining...
You can watch the full output in /laue/jamesh/showoff/Elves/hr/wedge1/logs/mosflm.log

      rmsd                 |xtal slip around|
frame preds mosaic  dist   beam   "Y"     phi  unit cell
   1  0.17    n/d  240.1   0.00   0.07  -0.03  38.189 57.602 147.314 90 90 90
   2  0.15    n/d  240.1   0.00   0.07  -0.04  38.189 57.602 147.314 90 90 90
   3  0.18    n/d  240.1   0.00   0.05  -0.04  38.189 57.602 147.314 90 90 90
  30  0.16    n/d  240.2   0.00   0.05  -0.02  38.189 57.602 147.314 90 90 90
  31  0.14    n/d  240.2   0.00   0.06   0.03  38.189 57.602 147.314 90 90 90
  32  0.15    n/d  240.1  -0.01   0.03   0.04  38.189 57.602 147.314 90 90 90
  59  0.14    n/d  240.2  -0.03   0.02   0.03  38.189 57.602 147.314 90 90 90
  60  0.12    n/d  240.1   0.00   0.04   0.04  38.189 57.602 147.314 90 90 90
  88  0.12    n/d  240.1  -0.04   0.03   0.06  38.189 57.602 147.314 90 90 90
  89  0.13    n/d  240.1  -0.05   0.03   0.05  38.189 57.602 147.314 90 90 90
  90  0.13    n/d  240.1  -0.02   0.03   0.03  38.189 57.602 147.314 90 90 90
 117  0.14    n/d  240.1  -0.01   0.02   0.04  38.189 57.602 147.314 90 90 90
 118  0.13    n/d  240.1   0.02   0.01   0.04  38.189 57.602 147.314 90 90 90
 119  0.14    n/d  240.1   0.01   0.01   0.07  38.189 57.602 147.314 90 90 90
 146  0.15    n/d  240.2   0.06  -0.06   0.01  38.189 57.602 147.314 90 90 90
 147  0.14    n/d  240.1   0.06  -0.06   0.01  38.189 57.602 147.314 90 90 90
 148  0.15    n/d  240.2   0.04  -0.03   0.00  38.189 57.602 147.314 90 90 90
 177  0.15    n/d  240.1   0.05  -0.17  -0.01  38.189 57.602 147.314 90 90 90
 178  0.16    n/d  240.1   0.05  -0.13   0.00  38.189 57.602 147.314 90 90 90
 179  0.04   0.33  240.2   0.05  -0.17   0.02  38.263 57.717 147.692 90 90 90
   1  0.16    n/d  240.6   0.00   0.09  -0.02  38.263 57.717 147.692 90 90 90
   2  0.15    n/d  240.6   0.00   0.07  -0.01  38.263 57.717 147.692 90 90 90
   3  0.17    n/d  240.6   0.00   0.07  -0.03  38.263 57.717 147.692 90 90 90
  30  0.16    n/d  240.7   0.01   0.06   0.00  38.263 57.717 147.692 90 90 90
  31  0.15    n/d  240.7   0.01   0.08   0.03  38.263 57.717 147.692 90 90 90
  32  0.14    n/d  240.6   0.00   0.06   0.03  38.263 57.717 147.692 90 90 90
  59  0.14    n/d  240.6  -0.02   0.03   0.04  38.263 57.717 147.692 90 90 90
  60  0.12    n/d  240.6   0.01   0.04   0.04  38.263 57.717 147.692 90 90 90
  61  0.14    n/d  240.6   0.01   0.04   0.04  38.263 57.717 147.692 90 90 90
  88  0.12    n/d  240.6  -0.02   0.03   0.04  38.263 57.717 147.692 90 90 90
  89  0.12    n/d  240.7  -0.04   0.03   0.04  38.263 57.717 147.692 90 90 90
  90  0.12    n/d  240.5   0.01   0.03   0.02  38.263 57.717 147.692 90 90 90
 117  0.14    n/d  240.7   0.02   0.01   0.02  38.263 57.717 147.692 90 90 90
 118  0.13    n/d  240.6   0.01   0.01   0.03  38.263 57.717 147.692 90 90 90
 119  0.14    n/d  240.6   0.02   0.01   0.06  38.263 57.717 147.692 90 90 90
 146  0.14    n/d  240.7   0.06  -0.06   0.01  38.263 57.717 147.692 90 90 90
 147  0.14    n/d  240.6   0.06  -0.06   0.02  38.263 57.717 147.692 90 90 90
 148  0.13    n/d  240.7   0.04  -0.03   0.01  38.263 57.717 147.692 90 90 90
 177  0.15    n/d  240.6   0.05  -0.17   0.01  38.263 57.717 147.692 90 90 90
 178  0.15    n/d  240.7   0.04  -0.10   0.01  38.263 57.717 147.692 90 90 90
 179  0.05   0.37  240.7   0.05  -0.17   0.02  38.265 57.731 147.688 90 90 90
Spots are 0.03 mm (rms) from their predictions.
(pixel is 0.0816 mm)

orientation changed 0.014°
mosaicity   changed 0.637°
unit cell   unchanged.
distance    changed 0.19%


Ding! Refinement was successful! 

Parameters are still drifting.  Refine again.
On the CCD, you should probably fix the X-Y scale.

Recommend:
mv postref.mat best.mat 
/laue/jamesh/showoff/Elves/Elfsheim/Wedger refine nointeg fixyscale

You could, at this point, do some refinement of your unit cell, etc, but I decided not to.

Do this now? [Yes/go auto]?
-> no

What's wrong? [nothing]?
-> <Enter>
Shall we update mosflm.com? [Y]?
-> <Enter>
mosflm.com is ready...
Cleaning up...

moving back to root directory...
cd .. 

Shall we process the rest of your data? [Yes]
-> <Enter>

Here is where Processer Elves are launched. If you want to use more than one CPU, you should exit here, and launch the Processer jobs by hand. Otherwise, The Elves main program will just run one Processer job.

Processer running as pid=19898 on machine
using "nice" cpu priority
ready to process:
hr/wedge1 peak/wedge1 inf/wedge1

First, a separate processing directory is set up for each wedge...

setting up ./hr/wedge1 with ./hr/wedge1/start ./index/mosflm.com and /data/semet/hr_1_001.img
setting up ./peak/wedge1 with ./peak/wedge1/start ./index/mosflm.com and /data/semet/peak_1_001.img
setting up ./inf/wedge1 with ./inf/wedge1/start ./index/mosflm.com and /data/semet/inf_1_001.img
finished

Then each, in turn is tested to make sure mosflm runs without errors

testing ./hr/wedge1 (/data/semet/hr_1_001.img)
testing ./peak/wedge1 (/data/semet/peak_1_001.img)
testing ./inf/wedge1 (/data/semet/inf_1_001.img)
finished

Next, each wedge is refined to convergence

processing ./hr/wedge1 (/data/semet/hr_1_001.img)
processing ./peak/wedge1 (/data/semet/peak_1_001.img)
processing ./inf/wedge1 (/data/semet/inf_1_001.img)
finished

Finally, an average unit cell is computed, and forced upon all wedges

repeating refinement with average cell:
38.3383 57.8834 148.091 90 90 90
processing ./hr/wedge1 (/data/semet/hr_1_001.img)
processing ./peak/wedge1 (/data/semet/peak_1_001.img)
processing ./inf/wedge1 (/data/semet/inf_1_001.img)
finished

Processer Elves will call Scaler Elves when they are done with all the wedges.

Scaler Elves v 0.9     Because you have better things to do.(TM)   James Holton 8-30-99

Reading parameters in wedges.txt
Getting protein sequence from wedges.txt

Organizing mtz data ... (Cntrl-C to end)
hr/wedge1/raw.mtz (180 frames)
inf/wedge1/raw.mtz (180 frames)
peak/wedge1/raw.mtz (180 frames)

Hang on...

Scaler elves will localscale and merge your data with

3 wavelengths:
0.97949 Å  "Finf"    <- reference
0.97911 Å  "Fpeak"
0.95373 Å  "Fhr"

 frames   from file            as wavelength
1 to 180  inf/wedge1/raw.mtz   Finf
1 to 180  peak/wedge1/raw.mtz  Fpeak
1 to 180  hr/wedge1/raw.mtz    Fhr

The following SCALA cards will be used:
SDCORR 1.3 0 0.03
INTENSITIES SCALE_PARTIALS 0.85 

Resolution : 2 Å - 54.233 Å
Space group: P212121
Unit Cell  : 38.34 57.88 148.09 90 90 90
        Vm : 2.25 (44% solvent)
Asymmetric unit contains:
1/4 of unit cell, 
36500 Da of protein, 
8 Selenium sites. (expected)

Everything look okay? [Yes]
-> 
Good. Don't go away.
setting up scripts in ./scripts/...
setting up solve in ./SOLVE
setting up shelx in ./SHELX
setting up X-PLOR in ./XPLOR
 
    Scaler Elves are now ready to start scaling your data.  Scripts
to run "solve" and "shelx" have been set up, but have no data
files yet.  Your raw data must be scaled and merged first.
    However, since scaling can take longer that you will probably 
want to hang around, there are a few questions we would like you
to answer now:

Would you like to:
1)  Have the Elves give you some merged data to look at ASAP.
2)  Use these scripts to process your data yourself, thankyouverymuch.
3)  Let the Elves edit the scripts, and improve parameters automatically.
Choose a strategy [3]
-> 3
Okay, you can go away now.
sorting raw data ...
prescaling reference set (Finf) ... (see ./logs/sorting.log)
roughly scaling everything to Finf.  (see ./logs/rough_scale.log)
Found 10 contiguous groups of scales.
updating run definitions to:
   frames   from file            as wavelength
  1 to  75  inf/wedge1/raw.mtz   Finf
 76 to  76  inf/wedge1/raw.mtz   Finf
 77 to  86  inf/wedge1/raw.mtz   Finf
 87 to  89  inf/wedge1/raw.mtz   Finf
 90 to 144  inf/wedge1/raw.mtz   Finf
145 to 180  inf/wedge1/raw.mtz   Finf
  1 to  99  peak/wedge1/raw.mtz  Fpeak
100 to 162  peak/wedge1/raw.mtz  Fpeak
163 to 180  peak/wedge1/raw.mtz  Fpeak
  1 to 180  hr/wedge1/raw.mtz    Fhr
roughly scaling everything to Finf.  (see ./logs/rough_scale.log)
switching to smooth scales.
smooth scaling everything to Finf.  (see ./logs/rough_scale.log)
localscaling: this can take a long time.  (see ./logs/localscale.log)

mergeing Fhr as ./mtz/Fhr.mtz ...  (see ./logs/merge_Fhr.log)
mergeing Fpeak as ./mtz/Fpeak.mtz ...  (see ./logs/merge_Fpeak.log)
mergeing Finf as ./mtz/Finf.mtz ...  (see ./logs/merge_Finf.log)

combining Fhr Fpeak Finf into ./mtz/all.mtz ... (see ./logs/scaleit.log)
wavelength Rmerge  Ranom I/sigma Complete Mult Wilson B
       Fhr  0.046  0.045  17.400    89.5%  5.5   49.377
      Finf  0.103  0.078   6.900    85.6%  4.8   48.024
     Fpeak  0.044  0.059  17.500    85.9%  5.4   50.132
adding Free-R flags ...

True completeness to 2.10 A:
Fhr:     88.80% complete,     <F>/<sig> =  35.59
Fpeak:   85.15% complete,     <F>/<sig> =  35.28
Finf:    81.03% complete,     <F>/<sig> =  12.52

look at end of ./logs/scaleit.log for Dano and Diso vs Finf
./mtz/all.mtz is ready for input into SHARP or mlphare.
./SOLVE/*.fmt are ready for input into SOLVE
./SHELX/*.hkl are ready for input into SHELX
./XPLOR/*.fobs are ready for input into XPLOR's mad_merge.inp
Thank you, drive through! 

Here I noticed that the "Finf" data had unusually poor statistics. Inspection of the frames revealed that this wavelength (which was collected last) was very badly decayed. In fact, the lattice seemed to be splitting in two. Let this be a lesson to all that MAD data should be collected in small wedges, while continuously rotating wavelengths.

At this point, the "Finf" data were removed from further analysis. Scaler Elves were then re-run manually, like this:

unix% ./Elfsheim/Scaler hr/*/raw.mtz peak/*/raw.mtz wedges.txt >&! Scaler.log &
unix% tail -f Scaler.log

After that finished (log similar to the one above), Processer Elves were run again, but this time only to run heavy-atom finding programs, no mosflm.

unix% ./Elfsheim/Processer solve >&! process2.log &
unix% tail -f process2.log

Now we need to find heavy metal sites. Since shelx is available, and fast, it gets run first. Because this crystal was orthorhombic, it is unclear as to which of the possible 8 combinations of screw/rotation axes are actually present in the cell symmetry, so all of them will be tried.

shelx will be run in each space group
running shelx in  P21212 at SHELX/P21212  CFOM = 0.1281
running shelx in P212121 at SHELX/P212121 CFOM = 0.1554
running shelx in   P2122 at SHELX/P2122   CFOM = 0.1335
running shelx in  P21221 at SHELX/P21221  CFOM = 0.1386
running shelx in   P2212 at SHELX/P2212   CFOM = 0.1515
running shelx in  P22121 at SHELX/P22121  CFOM = 0.1491
running shelx in    P222 at SHELX/P222    CFOM = 0.1138
running shelx in   P2221 at SHELX/P2221   CFOM = 0.1286

The CFOM numbers from shelx are the overall confidence in the directly calculated phases (lower is better). We could now start refining the sites found by shelx in each of these space groups by using Phaser Elves, possibly each on a different CPU. However, because solve was available, and licensed, Processer Elves moved on to trying solve in every space group, starting with the one with the best shelx score. If more than one copy of Processer Elves were running, they would have divided up the space groups between themselves.

running SOLVE in P222 at /laue/jamesh/keck/SOLVE/P222
P222         7 sites, fom = 0.45, Z-score = 10.63

running SOLVE in P2221 at /laue/jamesh/keck/SOLVE/P2221
P2221        8 sites, fom = 0.46, Z-score = 14.55

running SOLVE in P22121 at /laue/jamesh/keck/SOLVE/P22121
P22121       8 sites, fom = 0.47, Z-score = 22.23

running SOLVE in P21212 at /laue/jamesh/keck/SOLVE/P21212
P21212       8 sites, fom = 0.47, Z-score = 21.17

running SOLVE in P2122 at /laue/jamesh/keck/SOLVE/P2122
P2122        8 sites, fom = 0.45, Z-score = 16.01

running SOLVE in P21221 at /laue/jamesh/keck/SOLVE/P21221
P21221       8 sites, fom = 0.44, Z-score = 10.93

running SOLVE in P2212 at /laue/jamesh/keck/SOLVE/P2212
P2212        4 sites, fom = 0.43, Z-score = 12.15

running SOLVE in P212121 at /laue/jamesh/keck/SOLVE/P212121
P212121      8 sites, fom = 0.48, Z-score = 21.52
waiting for other instances to synch up...

SOLVE summary:
P212121   8 sites, fom = 0.48, Z-score = 21.52
P22121    8 sites, fom = 0.47, Z-score = 22.23
P21212    8 sites, fom = 0.47, Z-score = 21.17
P2221     8 sites, fom = 0.46, Z-score = 14.55
P222      7 sites, fom = 0.45, Z-score = 10.63
P2122     8 sites, fom = 0.45, Z-score = 16.01
P21221    8 sites, fom = 0.44, Z-score = 10.93
P2212     4 sites, fom = 0.43, Z-score = 12.15

Since P212121 had the best statistics from solve, it is the most likely space group. So, Phaser Elves was run on it first.

Phaser Elves v 0.9     Because trying everything isn't so hard.(TM)   James Holton 3-22-00

Evaluating Fs in mtz/all.mtz ..picked Fpeak as the best reference.
checking mtz/all.mtz

writing scripts/reindex.com
writing scripts/rrsps.com
writing scripts/more_atoms.com 
writing scripts/dm.com
getting atom coordinates from SOLVE/P212121/solve.status
refining occupancies only
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4891 (0.1769 - 0.7507)  best Phasing Power = 2.29 (a), 1.89 (c)
eliminated 1 bad atoms from scripts/mlphare.com
switching to XYZ refinement
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4909 (0.1775 - 0.7528)  best Phasing Power = 2.31 (a), 1.91 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4909 (0.1775 - 0.7526)  best Phasing Power = 2.31 (a), 1.92 (c)
enabling occupancy refinement
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4925 (0.1783 - 0.7543)  best Phasing Power = 2.34 (a), 1.95 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4925 (0.1783 - 0.7543)  best Phasing Power = 2.34 (a), 1.95 (c)
enabling B factor refinement
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4958 (0.1765 - 0.7490)  best Phasing Power = 2.31 (a), 1.91 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4958 (0.1765 - 0.7490)  best Phasing Power = 2.31 (a), 1.91 (c)
refinement has converged! 
running scripts/dm.com with 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, and 70% solvent for auto cycles each
25% : R_dm = 0.210 ( 4/4 )
30% : R_dm = 0.146 ( 4/4 )
35% : R_dm = 0.176 ( 4/4 )
40% : R_dm = 0.164 ( 4/4 )
45% : R_dm = 0.170 ( 6/6 )
50% : R_dm = 0.170 ( 7/7 )
55% : R_dm = 0.131 ( 3/3 )
60% : R_dm = 0.131 ( 4/4 )
65% : R_dm = 0.132 ( 6/6 )
70% : R_dm = 0.130 ( 10/10 )
copying scripts/mlphare.com to scripts/mlphare.com.best
copying mtz/dm70.mtz to mtz/dm_best.mtz
maps/best.map is a normalized ccp4 map from mtz/dm_best.mtz
maps/best.omap is an o version (dsn6) of maps/best.map
copying scripts/mlphare.com.best to scripts/mlphare.com.unflipped

At this point, 70% solvent gave the "best" real-space free residual from dm, so Phaser Elves made a backup of the solvent-flattened output file as mtz/dm_best.mtz, and made a map for you to look at in o. Phaser Elves tried a wide range of solvent contents because (for this run) I didn't specify what the solvent content was. If I had typed "30% 35% 40%" on the Phaser Elves command line, they would have just done those solvent contents.

However, there is still some ambiguity as to the hand of the site constellation. Also, because MAD can have positive or negative dispersive differences, the Elves don't know which is the correct one (I could have given Phaser Elves a pre-written mlphare script with the correct sign in the occupancies, but I decided not to make it easy for them...). So, now Phaser Elves will try negating the real (dispersive) occupancies and inverting the site constellation to the other hand, both separately, and at the same time...

inverting real occupancies 
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4929 (0.1751 - 0.7479)  best Phasing Power = 2.35 (a), 1.89 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4929 (0.1751 - 0.7479)  best Phasing Power = 2.35 (a), 1.89 (c)
refinement has converged! 
running scripts/dm.com with 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, and 70% solvent for auto cycles each
25% : R_dm = 0.168 ( 13/13 )
30% : R_dm = 0.170 ( 10/10 )
35% : R_dm = 0.141 ( 18/18 )
40% : R_dm = 0.211 ( 12/12 )
45% : R_dm = 0.219 ( 11/11 )
50% : R_dm = 0.270 ( 6/6 )
55% : R_dm = 0.217 ( 17/17 )
60% : R_dm = 0.222 ( 10/10 )
65% : R_dm = 0.230 ( 8/8 )
70% : R_dm = 0.218 ( 9/9 )
none of these were better than mtz/dm_best.mtz
scripts/mlphare.com.best is still the best mlphare script.

flipping all atoms through the origin 
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4929 (0.1751 - 0.7479)  best Phasing Power = 2.35 (a), 1.89 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4929 (0.1751 - 0.7479)  best Phasing Power = 2.35 (a), 1.89 (c)
refinement has converged! 
running scripts/dm.com with 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, and 70% solvent for auto cycles each
25% : R_dm = 0.149 ( 10/10 )
30% : R_dm = 0.123 ( 11/11 )
35% : R_dm = 0.169 ( 13/13 )
40% : R_dm = 0.149 ( 18/18 )
45% : R_dm = 0.160 ( 16/16 )
50% : R_dm = 0.182 ( 16/16 )
55% : R_dm = 0.176 ( 18/19 )
60% : R_dm = 0.207 ( 8/8 )
65% : R_dm = 0.174 ( 15/15 )
70% : R_dm = 0.202 ( 6/6 )
copying scripts/mlphare.com to scripts/mlphare.com.best
copying mtz/dm30.mtz to mtz/dm_best.mtz
maps/best.map is a normalized ccp4 map from mtz/dm_best.mtz
maps/best.omap is an o version (dsn6) of maps/best.map

inverting real occupancies 
and flipping all atoms through the origin 
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4958 (0.1765 - 0.7490)  best Phasing Power = 2.31 (a), 1.91 (c)
running scripts/mlphare.com into logs/mlphare.log .......... FOM = 0.4958 (0.1765 - 0.7490)  best Phasing Power = 2.31 (a), 1.91 (c)
refinement has converged! 
running scripts/dm.com with 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, and 70% solvent for auto cycles each
25% : R_dm = 0.185 ( 7/7 )
30% : R_dm = 0.193 ( 6/6 )
35% : R_dm = 0.191 ( 8/8 )
40% : R_dm = 0.214 ( 6/6 )
45% : R_dm = 0.212 ( 5/5 )
50% : R_dm = 0.187 ( 10/10 )
55% : R_dm = 0.186 ( 8/8 )
60% : R_dm = 0.206 ( 5/5 )
65% : R_dm = 0.175 ( 10/10 )
70% : R_dm = 0.165 ( 11/11 )
none of these were better than mtz/dm_best.mtz
scripts/mlphare.com.best is still the best mlphare script.
copying scripts/mlphare.com.best back to scripts/mlphare.com

Now, if these data had been in a handed space group (P41/P43, etc.), then it would still be possible to have the wrong hand, and Phaser Elves would now try reindexing the data to the opposite-handed space group, and then repeat the above procedure.

However, this crystal was orthorhombic, so there was no need to do this. The next step, therefore would be to calculate some difference Fouriers, and look for more sites.

Looking for more atoms in mtz/dm_best.mtz
that are not already within 0.67A
of the 8 (plus 1 bad) atoms in scripts/mlphare.com

resolution 1000 2
using FOMDM*F and PHIDM for difference Fouriers
fft will use F > 1*sigma(F) with |DANO| < 127.35 and |delta-F| < 102.45

DANOFhr DANOFpeak 
Fhr-Fpeak 
3 difference Fouriers
weighting maps...
combining maps..
49 peaks found > 3*sigma, (41 new).

making Pattersons
(DANOFhr)^2 (DANOFpeak)^2 
(Fhr-Fpeak)^2 
weighting Pattersons...
combining Pattersons..
checking 41 sites against average Patterson

all peaks and scores written to ./more_atoms.list
top 5 new sites: 
  sigma    X      Y      Z   is   dist from nearest neighbor
    4.6   0.819  0.809  0.839    3.07A      ATOM5   in DERIV 1 
    4.2   0.768  0.202  0.370    3.55A      ATOM3   in DERIV 1 
    3.2   0.766  0.736  0.834    3.29A      ATOM3   in DERIV 2 
    3.5   0.854  0.765  0.841    3.19A      ATOM5   in DERIV 1 
    3.4   0.864  0.715  0.847    5.42A      ATOM5   in DERIV 1 

These 5 sites were chosen not just because they were peaks in the average difference Fourier, but because they had relatively believable Patterson peaks, as well as cross-peaks to the current 8 sites.


At this point, maps/best.omap was inspected in o, and it looked like this:

Elves went on to try all the other solve solutions too, but it is clear from the above map that P212121 is the correct space group.

This map was obtained about three days after the original, raw data were collected at beamline 5.0.2 at the ALS. Most of this time was spent running solve.


Back to the Elves Page.


This page is not finished. It will never be finished, and neither will yours. Admit it.

James Holton <jamesh@ucxray.berkeley.edu>