| X-PLOR 98.1 |
The following section lists the major new features that have been added to the X-PLOR distribution since the release of X-PLOR 3.1.
All information related to atomic models (the number of atoms, number of bonds, etc.) is now dynamically allocated within the X-PLOR program. You no longer need to recompile X-PLOR when you work with very large atomic models. The computer memory is used more efficiently for small structures. Automated memory allocation for experimental data means that you do not need to define the number of structure factors or NMR restraints before they are read.
All of the data items that are contained in atom records in the standard PDB format are now maintained. This means, for example, that chain identifiers and disorder flags are read and preserved by the X-PLOR program. Note, however, that the atom selection mechanism does not yet allow picking on these items.
This version of the X-PLOR program contains the abnr (Adopted Basis Newton Raphson) minimization routine in addition to the Powell minimization routine. The abnr minimizer makes use of numerical second derivatives to improve convergence. Trials show that the abnr minimizer converges with lower energies than the Powell minimizer for typical crystallographic and NMR applications in a similar amount of a computer time. Furthermore, the abnr minimizer does not abandon its line search and stop prematurely as sometimes happens with the Powell minimizer.
This version of X-PLOR incorporates SGI's mp code to allow jobs to be distributed across several processors in multiprocessor machines. The acceleration achieved by the parallel processing depends on the size and type of the calculation, with the greatest gains achieved for crystallographic calculations for large structures. A trial crystallographic refinement involving the joint chemical and X-ray energy minimization of a 2026 atom structure with 17556 reflections (with an FFT grid of 96x108x120) sped up on a four processor machine by almost a factor of three.
1. Major update of all X-ray crystallographic tutorial files using new syntax
New features for X-ray crystal structure determination
The example scripts have been changed to reflect changes and additions to the X-PLOR command syntax, improved structure determination protocols, and the addition of standard libraries for space groups and atomic scattering factors.
Torsion angle dynamics, in which torsion groups within the molecule are kept completely rigid, increases the radius of convergence of crystallographic refinement by allowing simulated annealing at higher temperatures and with larger time steps (Rice and Brünger 1994). Torsion angle dynamics also has the benefit that it is more stable than conventional Cartesian dynamics and, since the number of degrees of freedom in the system is reduced, it is less prone to overfitting the data.
3. Efficient energy minimization script
As an alternative to crystallographic refinement with torsion angle dynamics, a powerful and highly automated multi-stage energy minimization script has been developed for the refinement of models with large positional errors. This script makes use of maximum likelihood methodology, the abnr energy minimization routine, and the Babinet bulk solvent scattering correction. This script uses X-PLOR macros to allow the automatic calculation of the weight, wa, for the crystallographic energy term and the optimal scale factor for the bulk solvent scattering correction.
X-PLOR is now able to carry out MAD phasing applications, including data scaling and merging, difference map calculations, refinement of the anomalous scatterer sites and phasing (Burling et al. 1996).
5. 
A-weighting for electron density maps with optional cross-validation
Improved electron density maps with reduced bias may be calculated using Aweights (Read 1986; Hodel et al. 1992) or cross-validated A weights. The calculation of these types of map facilitates model re-building.
6. Difference, anomalous difference, and <n>Fo-<m>Fc electron density maps
Example scripts are available for producing the types of electron density maps most commonly used in macromolecular crystallography.
A method
Scripts are included for estimating coordinate errors in macromolecular structures from the calculation of Luzzati (Luzzati 1952) and A (Read 1986; Kleywegt and Brünger 1996) plots.
8. Script files for molecular replacement with multiple molecules
Example scripts are provided for carrying out molecular replacement searches for the case where there are multiple molecules in the crystal asymmetric unit.
The X-PLOR program is able to pick water peaks (corresponding to ordered solvent molecules) from the region of the solvent that is close to the protein surface and write out the solvent co-ordinates. An example script is provided for this application.
A procedure (including an example script) for creating real-space bulk solvent density models (Jiang and Brünger 1994) and writing partial structure factors for these models is now included. With this procedure it should be possible to include all of the very low-resolution structure factor data in refinement and map calculations.
11. Reciprocal space bulk solvent refinement procedure
The Babinet bulk solvent scattering correction is now implemented in the X-PLOR program. In contrast to the real-space bulk solvent models, which are created for specific coordinate sets, this implementation provides a convenient and fast first-order correction for bulk solvent scattering that does not need to be manually updated during the refinement. This solvent scattering correction allows the inclusion of all very low resolution data in refinement calculations. An X-PLOR macro is provided for automatically determining the optimal scale factor for the bulk solvent scattering correction.
The command syntax is now expanded to explicitly support the direct rotation function. In this rotation function, the search molecule is placed in a crystal cell with the same dimensions as the unknown crystal and the rotation function is computed for the molecule systematically placed in different rotations (DeLano and Brünger 1995). This brute force rotation function is accurate but computationally expensive.
13. Phased translation function
The phased translation function is able to use the extra (phase) information that is available for locating the translational components of a trial model when, for example, phase information has been obtained from a heavy atom derivative. The algorithm searches for maximum correlation of observed and calculated electron densities (Read and Schierbeek 1988).
Two scripts have been provided to compute and record information on refined structures in formats that facilitate use of the automated deposition system at the Protein Data Bank. One script formats the information in a PDB coordinate file and the other script provides the information in the form of a mmCIF coordinate file.
Structure factor data may now be read directly by the X-PLOR program in a variety of formats. This capability simplifies the use of the X-PLOR program with newly processed data and with data sets that have been formatted for use by other programs.
The maximum likelihood targets for the refinement of macromolecular models provide much more convergent and reliable results than the conventional residual target, particularly when the model is incomplete and relatively inaccurate (Pannu and Read 1996; Adams et al. 1997). Since the computational cost of maximum likelihood refinement is little different from conventional refinement, the intensity-based maximum likelihood target is now standard (default). The example refinement scripts provided with this release have been altered to illustrate the use of the maximum likelihood targets.
The Andersen thermal coupling method (Andersen 1980) has been adapted for use with the simulated annealing code within X-PLOR. The Andersen thermal coupling method appears to give better local sampling than Berendsen thermal coupling and most structures refine to slightly better (~0.5%) Rfree values if the Andersen thermal coupling method is used in the final stages of simulated annealing refinement.
1. Time- and ensemble-averaged NOE distance restraints
New features for NMR structure determination
In structure determination by X-ray crystallography and solution NMR spectroscopy, experimental data are collected as time- and ensemble-averages. Thus, in principle, appropriate time- and ensemble-averaged models should be used. For refinement with time-averaged NOE distance restraints (Torda et al. 1989, 1990; Pearlman and Kollman 1991) the NOE restraint potential is changed so that distance restraints derived from NOE are applied to the time-average of each distance, rather than each instantaneous distance. With ensemble-averaging, an ensemble of conformers rather than one single structure is used to satisfy the experimental NMR data (Bonvin and Brünger 1996).
2. Cross validation of NOE distance, NOE intensity, and dihedral angle restraints
The idea of cross-validation for structure determinations involving NOE distance, NOE intensity, and dihedral angle restraints (Bonvin and Brünger 1995) is illustrated by the scripts used for ensemble averaging.
3. J-coupling and proton and carbon chemical shift refinement
Direct refinement against three-bond HN-CaH coupling constants (Garrett et al. 1994) is now available.
Direct proton chemical shift refinement (Kuszewski et al. 1995a, 1996) and direct secondary carbon chemical shift refinement (Kuszewski et al. 1995b, 1996) are also available and exemplified by tutorial scripts.
4. Ambiguous restraints and iterative assignments: ARIA
Methodologies for using ambiguous restraint references.htmlinformation to perform automated iterative peak assignment and structure determination (Nilges 1997) are implemented in this version of X-PLOR and a set of tutorial files is provided.
5. Structure calculation with torsion angle dynamics
Torsion angle dynamics provides an efficient method for NMR structure determination, with a high success rate when compared to conventional methods (Stein et al. 1997). An example script is provided for NMR structure determination using this protocol. The Stein et al. protocol involves the use of both torsion angle and Cartesian dynamics. A sample protocol is also provided for NMR structure determination that uses only torsion angle dynamics.
The torsion angle algorithm (Rice and Brünger 1994) has been recorded in more computer-efficient form giving a saving (depending on the size of the structure) of ~20% on the total run time for these protocols.
6. Fast direct NOE intensity refinement
A new methodology for computing NOE intensities and gradients has been incorporated into this version of X-PLOR. This new technique is based on the Matrix Doubling and Gaussian Quadrature approximation (Yip 1993). The increase in speed over methods available in X-PLOR 3.851 spans one to two orders of magnitude.
7. Engh-Huber parameters for NMR structure determination
A parameter set based on the Engh-Huber parameters for X-ray refinement, but including all hydrogen atoms for NMR structure determination, is included in this release of X-PLOR.
Methodology for direct refinement using information on cross-correlated dipole-dipole rates is implemented in this version of X-PLOR.
X-PLOR source code and release tree
X-PLOR program
The source code for the X-PLOR program is now written in FORTRAN90 and was compiled using standard UNIX `make' facilities. The conversion of the X-PLOR source code to FORTRAN90 has facilitated the introduction of full dynamic memory allocation for atomic data and allowed greater internal modularity in the program. Re-engineering the source code in this way should allow more rapid application development in the future and the use of the most effective compilers. The source code for the X-PLOR program is not normally supplied with the release.
> setenv MP_SET_NUMTHREADS 4
will cause execution of the X-PLOR program from the same UNIX window to use four threads.X-PLOR users on SGI platforms are required to have a FORTRAN90 7.2.1 runtime environment including complib 3.0 (or later) installed on their machines. In addition, the runtime environment should support mp directives. This environment should be available on newer machines but older computers may require SGI patches to be installed. To learn about the patches needed to run this version of X-PLOR, refer to the document describing the patches required to run Insight II on SGI systems. See the Insight II release notes at the MSI web site (http://www.msi.com).
On most platforms the environment variables used to access these directories will be set up automatically.
nmrlib contains standard library data for NMR applications.
test contains a set of input scripts used for testing and diagnosing problems with the X-PLOR program. The test directory is further divided into a set of subdirectories according to the type of application.
toppar contains topology and parameter files needed to carry out calculations with atomic structures.
tutorial contains example scripts that can be used as templates for calculations with the X-PLOR program. This directory is further divided into subdirectories according to the type of application. Note that example data and output files are not supplied with this release.
xtallib contains library data that is used in crystallographic calculations.
xtalmacro contains a set of macros that can be called from X-PLOR scripts to carry out specific tasks.