The new skeletonization algorithm has been finished. It is triggered as before "GEN_SKEL" on page 10 (MAIN_MENU:map_mask.html). The new skeleton gives clearer traces of chains and side chains and is as such a better input for the initial hybridization of the model ("TRC_HYBR") as the beginning of the auto building pipeline. Naturally also the outcome of the "TRC_HYBR" has been thereby improved...
Split of "NCAC" into two buttons
It turned out that the output model from trace of main chain "TRC_NCAC" after some editing ("DELE_BON,"MAKE_BON","DELE_ATO", "MAKE_ATO") was often worse than the input. Therefore the function has been splitted in two:"TRY_NCAC"and"ACC_NCAC". The"TRY_NCAC" only tries to trace the main chain, whereas the trial model is accepted as the real model only after the"ACC_NCAC"has been clicked. Thereby the input model to"TRY_NCAC"can be edited manually and is not changed by the function itself"not changed by the function itself.
As a result of visit in Nenad Bans lab, the libraries for DNA and RNA have been updated. For reference use:
G. Parkinson, J. Vojtechovsky, L. Clowney, A.T. Brunger, H.M. Berman, New Parameters for the Refinement of Nucleic Acid Containing Structures, Acta Cryst. D, 52, 57-64 (1996).
The January release had some numerical problems with the maximim-likelihood refinement target function and sigma A map calculations. They have been fixed.
The January release contained a prerelease of the new skeletonization algorithm, which was unfinished and contained a couple of bugs. The algorithm has been finished and the bugs were exterminated.
Maximum likelihood has found a way into MAIN
With the help from the crew of Sasha Urzhumtsev from Nancy (Pavel Afonine, Andrej Fokine), who have come to Ljubljana for a ten days visit at the end of November. We introduced a few additional target functions for crystal structure refinement and map calculations in MAIN (least square approximation of the maximum likelihood refinement target function, maximum likelihood refinement target function, bulk solvent correction for refinement and map calculations and sigma A map calculation). The computational parts of this code were in addition to Sasha, Pavel and Andrej, partially written and developed also by Volodja Lunin, whereas I have integrated them into MAIN and written the "true" maximum likelihood refinement function based on Pavel derivation.
There are still some problems with Maximum likelihood B-value refinement with occasional "INFINITY" results and gradients as well as shift of average B-values - I hope work this out soon.
Depp page 10 is altered. "MAP_ATOM" and "KICK_STUF" menus have merged and split into new functional blocks for refinement MAIN_MENU:refine.html and map calculation MAIN_MENU:map_calc.html.
Also MAIN_DOC:calc_map/calc_map.html and MAIN_DOC:refine/refine.html have been accordingly modified.
For the maximum likelihood refinement targets use of R-free is unfortunately mandatory.
Bulk solvent correction is included in MAIN_MENU:map_calc.html.
The B-factor refinement has been now all the way equipped with double precision variables and which did indeed improve the stability and convergence of the B-factor refinement.
New rigid body minimizer interface, which includes also bonding terms, has been added. Currently it is applied in residue fitting ("RIG_RESI") MAIN_MENU:auto_stuff.html, but will be extended into peptide fitting and chain trace procedures.
New skeletonisation algorithm is on a way. Not finished yet, but since it seems quite useful, it is already included in the current release.
Model building seemed to be the worst part of the manual, so I have particularly filled the gap for new structures building ("MAIN_DOC/build/new_struc.txt").
Also the "MAIN_DOC/analyze/analysis.txt" has been completely rewritten.
The manual is not finished but I believe that it is better than it used to be.
NEWS: AUTOMATION is ON LINE !!!!!!!
By the end of June new release of main config tools will be released (after it goes through one week of testing during the annual MAIN course in Martinsried). All create tools have been now moved to Perl and a few features added to make MAIN configuration and model building easier
Auto-start with the .main file
Your local ".main" loads your "main" session by loading the data.
Extract information from MTZ files directly
"use_mtz.pl" extracts information about cell constants, space group and symmetry operators, resolution and structure factor columns. And you can already rotate your initial map on the screen. Two command lines are sufficient:
create_main_config.pl -u "my.mtz" --doit mainps
There is also no more editing of "mtz_2_main" and phase combination macros. They are generated automatically. Labels for FOBS, PHASES, FOM and HL-coefficients are chosen by the principle first on the list. If you are unhappy with defaults you can provide your choice via "use_mtz.pl" parameters.
AUTOMATED model building and density modification from the beginning: "RoH_jobs.cmds" generation is evolving: automated start of density modification and model building in the first maps.
OpenGL drivers which directly address the graphics are mandatory. Requirement for work are full filled already by most bottom line graphic cards.
NVIDIA graphic cards: good for mono work, but no crystal eyes stereo is supported for LINUX. GeForce3 and better achieve excellent performance.
MAIN supports crystal eyes stereo on any hardware assuming that Xserver supports it.
ATI FIRE GL2 and GL4 provides good stereo graphics. The cheaper one GL2 is already OK.
MAIN on LINUX PCs: dials - dialbox
Use of Silicon Graphics dials is supported thanks to the driver provided by Francois-Regis Colin,<fcolin@cenatls.cena.dgac.fr> and Frederic Lepied,<Fredric.Lepied@sugix.frmug.org>. France and adapted to XFree86 by Joe Krahn<krahn@niehs.nih.gov>.
The driver is provided also on MAIN FTP server FTP://stef.ijs.si/pub/.
Only XFree86 4.2 relese supports all 8 dials (earlier XFree releases may have bugs). Official ATI Xserver is still XFree 4.1 based - so dials do not work yet.
MAIN does not use CPU when idle anymore. An "if" condition for auto generation of events was turned inside out. Sorry ....
All executables from may 2002 onwards are OK.
There was a problem using FFT in cubic space group p41_3_2, which has been fixed.
Maps used for refinement have been changed to real. Old and new macros are not compatible any longer.