Sushi 3.4.0 - release notes

• Introduction
• Installation and upgrade
• List of changes (new features, improvements, fixes)

Introduction

This release has been tested on

• Linux
  1. ia32: Pentium 4, Xeon, Athlon etc
  2. x86_64: AMD64, Opteron
  3. ia64: Itanium2
• Mac OsX 10.4.X (aka Tiger)
• SGI/Irix
• HP Tru64 4.0F/5.X (note: SHELXC/SHELXD will not work on 4.0X systems)

IMPORTANT :  this release requires a new set of licence keys (so your existing .licence file will not work): please request licence renewal, additional licence keys, or a complete new licence

Installation

1. Download all necessary packages. This requires a password - which should have been sent to you at the time of licence acceptance (but it is changed regularly, so see also the password reminder page).

2. follow the installation instructions, especially the typical installation scenarios. E.g.:

   % cd /where/ever/sharp
   % ./installSHARP -F
         

1. start CCP4i
2. open "System Administration" (at the right side)
3. select "Install Tasks"
4. for the "Task archive", browse to the installation directory of SHARP/autoSHARP and select the file autoSHARP_CCP4i.tar.gz
5. click "Apply" and then quit/restart ccp4i
6. there should now be a task autoSHARP in the "Experimental Phasing" module and also in the "Program list" (by default new tasks are installed at the end of the list).

Upgrade

1. Download all new packages. This requires a password - which should have been sent to you at the time of licence acceptance (but it is changed regularly, so see also the password reminder page).

2. unpack files and update your configuration (in the so-called $BDG_home directory):

   % cd /where/ever/sharp
   
   % source ./setup.csh              # for csh/tcsh
     - or -
   % . ./setup.sh                    # for sh/bash/ksh
   
   % adm/bin/kill_server
   % ./installSHARP -U
         

3. see also the Sushi 3.4.X and general upgrade notes.

Changes since version 3.2.3

1. autoSHARP - new features:
   • The detection of the heavy atom substructure can now also be done with the SHELXC/SHELXD programs by G. Sheldrick. This is now the recommended way, although some internal parameters might still benefit from fine-tuning. As an alternative, the RANTAN procedure is also still available.

   • The user can now decide, if autoSHARP should use some short-cuts (hopefully speeding up the calculation but at the cost of less accuracy) or use the full set of features (being as accurate as possible, but maybe a little bit slower). This is done as a Speed/Accuracy setting in the interface. The main shortcuts that will speed up a calculation are done at the following steps:
     • data analysis and scaling (fewer outlier checks, so some problems with the given data might be missed);
     • heavy atom detection with SHELXD (fewer trials, which might miss the correct solution if the data or substructure signal isn't of high quality);
     • density modification (larger steps and fewer cycles during optimisation of solvent content, so that the correct number of molecules might be missed when a large number of NCS-related molecules is present);
     • automatic building using ARP/wARP (more frequent building steps and fewer cycles, possibly resulting in fewer residues being built or not all chains docked into sequence);

   • A new CCP4i interface for running autoSHARP is available: this has been tested with CCP4 5.0.2. The program output still uses HTML, so it is advisable to view these with a browser (Mozilla, Firefox, Internet Explorer ...). Please note: this mode of operation does not require a running httpd.

   • Support for viewing results (maps and models) with Coot by P. Emsley.

   • When estimating the most likely number of molecules per in the asymmetric unit (before any heavy atoms have been found and before an optimised solvent content is available), the Matthews coefficient probabilities by Kantardjieff & Rupp (Protein Science 12, 1865) are implemented.

   • During the analysis of residual maps (to update and check the current heavy atom model), an automatic decision based on SHARP residual maps (LLG maps) is done for the refinement of anomalous scattering factors f'/f''.

   • When deciding which is the correct hand (and also when the solvent content is optimised), a new score is used for density modification: CC/Contrast (CC is the correlation between E² of density-modified map and observed data and show how well the density modification fits with the observed data; Contrast is the ratio between the standard deviation of density in the solvent region and the standard deviation of density in the 'protein' region - this is a measure how macromolecule-like the modified density appears).

   • When several datasets are given (e.g. for a MAD experiment) and the space group allows it, all datasets are checked for alternate indexing to give the best agreement to the first dataset. These happens for space groups: P4 (75), P41 (76), P42 (77), P43 (78), I4 (79), I41 (80), P3 (143), P31 (144), P32 (145), H3 (146), P312 (149), P3112 (151), P3212 (153), P321 (150), P3121 (152), P3221 (154), H32 (155), P6 (168), P61 (169), P65 (170), P62 (171), P64 (172), P63 (173), P23 (195), F23 (196), I23 (197), P213 (198) and I213 (199). Also, the consistency of screw-axis components in all datasets is checked.

   • New reflection file formats are supported: unmerged SCALEPACK and SHELX HKLF 4 format files.

   • During the analysis of log-likelihood gradient (LLG) maps from SHARP (residual maps): already refined heavy atoms sites might be deleted if the standard uncertainty of their heavy atom occupancy parameter is too large.

   • To facilitate the connection of external helpers (to view results using Mozilla/Firefox directly from the HTML log files with e.g. O, Coot, Xfit etc), a new script is provided (add2rdf.sh) that will modify each users browser configuration (as long as it is found in ~/.firefox or ~/.mozilla/firefox/). A similar script for Konqueror (add2konq.sh) is available too - although not as well tested.

   • An autoSHARP paper should be out quite soon:

     Vonrhein, C, Blanc, E., Roversi, P and Bricogne, G. (2005). Automated structure solution with autoSHARP. In "Crystallographic Methods", S Doublie, Ed., Humana Press, Totowa, NJ, submitted.

     Steps that involve use of external programs will give clear reference to publications: make sure to cite all programs used.

2. autoSHARP - improvements:
   • The handling of off-origin peaks in native Patterson has been improved.

   • Some unnecessary warnings in initial checks are now avoided - to prevent some confusion.

   • A better automatic analysis of MAD wavelengths is done. Since the heavy atom detection with SHELXC/SHELXD require a correct identification of each wavelength by name, it is recommended to use some 'standard' convention as identifiers of the different wavelengths (as a fall-back plan for autoSHARP): peak, infl, hrem, lrem.

   • Better residual map analysis: avoid adding C_Sites with same G-Site and keep track of already removed sites. This should cut down on the number of (sometimes unnecessary) cycles between refinement of heavy atom model (SHARP) and the analysis/update.
   • During density modification, extreme values of solvent content are now avoided: this should stabilise this step in cases of very low (around 20%) or very high (above 70%) solvent content.

   • The interface to ARP/wARP for the automatic building of models into density modified maps has been updated to 6.1.1 logic. This should now be very similar to the currently distributed ARP/wARP 'warpNtrace' mode. Therefore, it might be beneficial to use the distributed ARP/wARP interface with data out of SHARP/autoSHARP. However, some settings in the ARP/wARP interface within autoSHARP have been optimised to work best with data coming out of SHARP/autoSHARP - so it is worth trying both!

   • If the resolution is not high enough for doing complete auto-building with ARP/wARP, the helix-building module of ARP/wARP will be used for getting at least part of an initial model (if there are helices).

   • The description of spherical cluster refinement (SPHCLUSTER keyword) has been improved in the distributed documentation. An example Sharp INput (SIN) file is available in $BDG_home/docs/sharp/doc.

   • Reflection files don't need to have an extension of *.data.mtz (or *.data.sca, *.data.hkl) anymore: a plain *.mtz file is sufficient.

   • During scaling of different datasets relative to each other: the low resolution analysis and adjustment of a low resolution cut-off have been stabilised.

   • Some more checks regarding the sequence file have been added.

   • If the data was processed with some resolutions shells excluded due to ice-rings on the images, potential problems during scaling/analysis are caught.

   • Job submission:
     • problems should be analysed better
     • support for PBS/OpenPBS
     • more meaningful job names for queueing systems

3. autoSHARP - fixed bugs:
   • switching hand for P3121/P3221

   • interface to ARP/wARP: inclusion of heavy atoms

   • adjustment of no. of molecules and no. of heavy atoms based on Matthews coefficient now takes user-given information better into account

   • better handling of CCP4 4.2.2/5.0.2 differences in MTZ file header (regarding space group and Project/Crystal/Wavelength (Dataset ID) information)

Changes since version 3.2.0

Changes since version 3.0.15

• compatible with CCP4 5.0.2

  Important :

  due to an error in the CCP4 5.0 MTZ library, the official CCP4 5.0.2 version of the DM program will not work correctly. A version of CCP4 5.0.2 that contains a fix (together with several others not yet in the official CCP4 5.0.X release(s)) is available from our site.

• lots and lots of small changes in autoSHARP (too many to list them all) that should improve
  • analysis of the various datasets
  • information a user gets presented (that helps making decisions)
  • heavy atom detection (several small bugs of the previous version have been corrected - these usually only occurred under special circumstances)

• added additional, basic health checks whenever a user connects to the SHARP Control Panel - so possible problems might get spotted early on.

• a new section of tutorials/howto documents to help a user getting started - see the relevant link from the SHARP Control Panel.

• added more checks at beginning of program usage - to make it easier to spot potential installation problems.