New Features in TNT

Table of Contents

• Release 1
• Release 2A
• Release 2B
• Release 3
• Release 3A
• Release 3B
• Release 3C
• Release 3D
• Release 3E
• Release 4B
• Release 4C
• Release 5A
• Release 5B
• Release 5C
• Release 5D
• Release 5E
• Release 5F

Release 1

• DIFFPLOT -- Plots the differences between two coordinate sets.
• BPLOT -- Plots the main chain B values.
• MCONT -- Using data PUNCHed by NEWGEO this program contours a TenEyck density map and can superimpose atomic labels and bonds.
• DIRDOT -- Calculates the angle between two direction file shifts in refinement space.
• MODEL -- Calculates a model electron density map from a set of coordinates for use in a FFT program.
• CONVERT -- Converts ATOMx formats to Diamond or AMSON or back again.

FUTURE GOALS

Work is also progressing toward a more rational description of multichain proteins. In the present package and most other packages if one has a structure with two polypeptides of identical sequence each corresponding residue must have different names and the amino acid sequence must be given twice. A method to avoid this has been devised and is being written into the next release of this package.

I am considering how I can handle curvatures within the package. I have some ideas but I have not worked out the details of a implementation. 

Release 2A

I have prepared release 2a and am sending it off. This release has several improvements over release 2 and all the errors reported Release 2 have been corrected. The FFT subdirectory programs have been completely reworked with several new space groups added. They all now conform to a uniform standard of file units and input parameters. I had hoped to have more Fourier synthesis programs ready but they have proved more difficult than I had expected at first. The program EXPANDER is very useful for expanding Fourier coefficients to lower symmetry space groups, especially to P1 but it will also expand to monoclinic and orthorhombic space groups. This additional functionality is not useful until some FS programs exist for those space groups.

I have at long last fixed the problem I was having with temperature factor gradient calculation. This correction is covered in the write-up in the sections concerning the new BLUR card for ADERIV. BDERIV is now no longer useful but has been maintained in the release "just in case". 

Release 2B

Here is release 2b. The major difference between 2a and 2b is that I have fixed a great number of the errors in Release 2A. I have added two new FFT programs: P2YSF and P2ZSF, and worked a little on MCONT (though a write-up for that program still does not exist.). 

Release 3

This is release 3. There have been many changes between release 2b and 3. A new structure factor calculation program, which is space group general, has been added. A great many modifications, corrections, and additions have been made, the number being too great to allow a complete list. All known errors have been corrected. 

Release 3A

This is release 3A. This release incorporates corrections for all known errors. A new version of GENERALSF has been included. This version reads the coordinate file directly, and eliminates the need to write the map file to disk. This programs expects a different set of input than the old version and command files will have to be modified.

This version also has a modification to NEWGEO which allows that program to process EXCLUDE cards with wild cards much faster. Using wild cards to specify an entire chain could take an enormous amount of time before but the new code is very fast. 

Release 3B

There are few differences between 3A and 3B. Other than a few minor bugs which have been eliminated the main difference is the inclusion of a number of new Fourier synthesis programs.

The default behavior of EXPANDER has been changed. Before it always chose an asymmetric volume with L positive. This was proper for Y sectioning FFT's, but those programs are no longer used. Now the K direction is always restricted to positive values and the user can choose to restrict H or L. 

Release 3C

New to this release is the incorporation of the real space refinement module, and the documentation of the BCORREL and phase refinement modules. These new modules have not been extensively tested and may incorporate bugs.

Also several types of VAX specific code have been changed to FORTRAN-77. There are still quite a few things which a compiler insistent on FORTRAN-77 will choke on, such as many variables and subroutines with names longer than six letters.

RFACTOR can now compensate for the lack of a solvent continuum in the protein structure model. This is effected with the SOLVENT card. The method used has some quirks and will behave badly in some circumstances.

Release 3D

OVERLAY can now read an arbitrary transformation via the MATRIX and TRANSLATION cards.

SHIFT now prints a short summary of all the contributing modules in the long loop.

ADERIV requires that a BLUR card be given whenever gradients are to be calculated.

CONVERT now produces CHAIN cards when a PUNCH SEQUENCE command is given. It does not put linkage information on these cards.

Release 3E

All usage of BYTE arrays for data input have been eliminated. This change improves the adherence of the source code to FORTRAN- 77 standards. Most of the remaining non-standard elements of the source code are inconvenient but not difficult to change.

All of the space-group specific Fourier synthesis programs have been replaced by GENERALFS. The manual has been modified to describe this new program.

Release 4B

TNT in general:

• Rotation function refinement
• Rotation function maps
• Use of MAKE for recompiling
• Read input from terminal instead of FOR001
• Major programs operate on VMS and E&S UNIX

• No longer requires Weights
• No longer requires scale factors from RFACTOR
• Removed CURVATURE command

• Difference Paterson coefficients with IO-IC
• Added the PUNCH COEFFICIENT command
• Rewrote the FUNCTION command
• Added GRADIENT and CURVATURE cards
• Both write out proper coefficients for FFTDERIV
• CURVATURE cards writes out curvatures
• New curvature code more accurate
• New curvature code can sigma weight
• Cleaned up REPORT ANISO code: Now produces standard Bij's

• Added COPY command card

• Added checking of incoming parameters against RANGE cards
• (Only if those parameters will be shifted)

• Added the JIGGLE command
• Removed the ADDB card and created ADD

Release 4C

General:

• Conjugate direction method added.

• HKL_FILES utility created
• ADDHKL, PAKTOHKL removed from package

• Symmetry handling corrected.
• Curvature still does not work.

• Presort now optional

• COMBINE cards cannot be used with gradient over curvature or conjugate directions.

Release 5A

All TNT

• Removed all INTEGER*2 declarations. This allows the maximum number of atoms to be raised above the 32K boundary. The current maximum is 80,000 atoms.
• The NOTE card has been created. This card contains an unechoed remark. In effect it causes no action to be performed. Your can use it to include quite remarks about the contents of a data file.

• Added a PUNCH <> BLOCKDATA command. This command writes the FORMFACTOR card information in a BLOCK DATA format suitable for compiling with the FFT programs. Also added PUNCH <> ATOMx command for compatibility.
• Program has been removed from package. All functions subsumed into RFACTOR.

• Real space refinement works once again. The procedure is much simpler than before. All specific functions are performed by RFACTOR in a fashion directly analogous to reciprocal space refinement.
• Program now can calculate and PUNCH maps. The functions of ADERIV and FFTDERIV have been incorporated and those programs have been removed from the package.

• New program created which replaces GENERALSF, GENERALFS, MAPREADSF, MODEL, HKL_FILES, ADD_HKL, and includes additional functionality. Eliminates the need for EXPANDER. The replaced programs have been removed from the package. EXPANDER remains.

• NEWGEO has been renamed GEOMETRY
• The REPORT GEOMETRY command now produces meaningful statistics for torsion and pseudorotation angles. They are split into a separate class for each legal value and the mean and sigma are produced for each.

• Moved to the utility area (TNT_UTIL:)

Release 5B

ALL TNT

• Numerous problems were discovered with the Unix version upon porting to the Convex. The corrections have been incorporated into the current version.
• Makefile macros are now listed for Sun-4 and DECstation workstations in the MAKEFILE.MIPS file. TNT still will not function on a Stardent.

• CHAIN card linkages have been redefined to mean that the linkage is to the nearest symmetry image of the target.
• The LINK card has been created to allow linkages to symmetry related molecules in single chain structures.

• The data request code has been modified to allow the processing of command verbs. This allows the calling program to perform almost any transformation of the data using this mechanism.

Release 5C

FOURIER

• Added CORRECT source to PUNCH command. This modifier causes the data to be moved to the correct asymmetric unit and sorted in the proper order before being dumped to disk.

• Added program to package. This program will calculate Wang type envelopes and produce solvent flattened maps.

• Added the ability to read and write DSN2 format files.

• Can write maps in DSN6 format.

• Added SWAP_PHASE coefficient type to replace the functionality of SWAP_HKL. The coefficients produced contain the amplitude from the FO file (scaled to the FC's) and the phase from FC.

• Added more intelligent B correlation restraints to this program. Also devised a library of standard values. BCORREL is now obsolete.

Release 5D

All Programs

• Expanded the power of the code which reads the input. One can now continue a card over several line by having either a "-" or "\" on the end of the line. One may also declare the remaining portion of the line a comment with a "!" or "%".

• When reading PDB files Convert will now convert the C-terminus from a PDB style one to TNT's CTERM linked residue. In addition Convert will read the SSBRIDGE cards in a PDB file.

• These programs will now write correct DSN6 files on big endian machines.

• A program for constraining and restraining a model to noncrystallographic symmetry has been included in the package. It will refine most such cases.

Release 5E

All Programs

• The continuation character "-" must now be preceded by a blank. This change was made to reduce the confusion resulting from names which end in this character, like "CL-".
• The TRANSFORMATION statement allows general linear transformations to be defined and later used with the APPLY command. A transformation can be expressed as a matrix, the rotation about an axis, Euler angles, or an equivalent position. Translations can be applied as well. The transformation can be expressed in the coordinate systems - TNT orthogonal, PDB orthogonal, or fractional.

• Fourier will now find peaks in a map. This is done by PUNCHing coordinates with a map as a SOURCE. An attempt to estimate the B value a water molecule would have to have to produce a peak of that height is made. For this calculation to be accurate the map's density values should be in e/A^3, which means that a difference map must be multiplied by 2.

• Most operations involved in refinement can be performed using canned scripts. A .tnt file is required. The new commands are:
• tnt -- Runs several cycles of refinement
• geometry -- Screens for the worst geometry
• rfactor -- Prints the R-factor and scaling
• correct -- Moves the Fobs to the proper asymmetric unit
• diffcor -- Compares two coordinate files
• pekpik -- Finds the highest peaks, makes DSN2 file
• make_maps -- Makes Fo-Fc and 2Fo-Fc DSN6 maps
• to_pdb -- Converts the coordinate file to PDB
• from_pdb -- Converts a PDB file to TNT format
• to_dsn2 -- Converts the coordinate file to DSN2

Release 5F

 CONVERT

• Convert can reformat TNT style geometry restraints, expressed on GEOMETRY cards, to Frodo and O's REFI format.  While much of the work is done automatically you will likely need to fine tune the result.  A document has been written to describe the process.
• Convert will now read PDB files with multiple models.  This usually only occurs for NMR models.  Convert will construct a chain name which is a combination of the PDB chain name and the model number.  For example chain C of the third model becomes C_3.
• An option has been added to the PDB reading subroutine which allows all hydrogen atoms to be ignored.  This option is NO_HYD.
• When PUNCHing PDB files Convert will convert a TNT style C-terminus to a PDB style.  This involves moving the extra oxygen into the final amino acid and naming it OXT.  This conversion does not work with statically disordered C-termini.
• When PUNCHing PDB files residues will be written in sequence order.  All peptides and nucleic acids will be written from the beginning to the end.  This does not mean that the polypeptide fragments will come out in the order you expect, only their amino acids are ordered.  In the past TNT would often write amino acids out of sequence since TNT does not care about the order of atom records.
• Convert now interprets and writes TER cards when working with PDB format files.  This improves the reliability of the sequence file generated from a PDB file, if the PDB file itself contains TER cards, and most do not.
• The ADD and JIGGLE commands now will use the RANGE statement to generate upper and lower bounds to B and OCC parameters.  These defaults can be overridden by adding a RANGE modifier to either command.

•   This script now sets an option which causes Convert to omit all hydrogen atoms from the new coordinate file.

• Added the PUNCH SCALE command.  This will write a SET card to a file which can be read back into the program in another context.  This is useful for calculating R-free.
• You may now PUNCH maps with Fourier coefficient types (FO-FC)**2, and IO-IC directly.  In the past you were required to write these coefficients to disk and transform them with Fourier by explicitly specifying the Patterson space group.  Rfactor now knows how the space group is changed during the recombination of crystallographic data.
• Anisotropic scaling between Fo and Fc is now be performed during refinement and map calculation.  If you have refined a model without anisotropic scaling the initial scale factors calculated by Rfactor will be underestimated in magnitude.  As you refine they will become larger.
• The ACCEPT statement will now allow defaults to be used.  If you only put one number on a line that expects two that number will be presumed to be the new upper limit and the default value used for the lower limit.  The default is not to impose a limit at all.
• The determination of scale factors relating Fobs and Fcalc has been made much more robust.  This step used to fail often and ask for the user to create SET statements.  Even though the program can always calculate values for the scale factors the actual values may be physically unreasonable.  If they are bad enough the program will terminate, demanding more SET statements.  Even if the program runs you must still be vidgulent   and add SETs if the scale factors are poor.

• You may now read a set of Fourier coefficients and PUNCH out a Patterson map.  This is done with the PATTERSON modifier to the PUNCH command.  Just use PATTERSON in place of SOURCE.

• Added the CLEAR card which allows previous information to be ignored.  This is useful for overriding TNT's default geometry library, or scattering factors.  You can also clear the space group or the atoms, if the need presents itself.
• The ASSUME RESIDUE_TYPE statement has been created.  This statement allows you to define the default residue type for a residue which contains all the atoms in a list of atom names, but no extras.  Since this is the equivalent of a RESIDUE statement without any links, it is only useful for solvent residues. However, it does allow you to freely create and delete solvent without being forced to edit the sequence file. Built into all programs is the ASSUME statement which defines any residue which contains only an OH atom to be of type HOH.  No more sequence file entries for water molecules!

• The TRANSFORMATION card has been extended to include a SHIFT B and SHIFT OCC term.  This allows refinement of structures where the ncs copies vary in average B factor or occupancy.

• This command has been changed to produce PDB format files instead of DSN2.  To regain the old results use the new Pdb2dsn2 command.  (The Unix version of TNT has always behaved in this fashion.)

• The minimization method used by default has been upgraded to "atomic block diagonal preconditioned conjugate gradient".  Each module will produce a 5x5 block for each atom and pass it to Shift.  Shift will perform a "singular value decomposition" of each atom's block, isolate the nonsingular subspace, construct the pseudoinverse, and calculate the primary shift by multiplying the pseudoinverse and the gradient.  The primary shift is then modified via the preconditioned conjugate gradient procedure to generate the actual shift direction.  The biggest improvements arising from this method are the proper handling of atoms on special positions, without user intervention, and better optimization in the presence of ncs restraints.  The Geometry module does not produce blocks, only diagonals, because block diagonal degrades convergence for that module.

• The CORRECT source in the PUNCH command of Fourier, and therefore the Correct shell command itself, will now filter out the "notches" in the reciprocal space asymmetric unit.  While Correct previously allowed a small number of redundant reflections into the HKL file only unique reflections are now allowed to pass.

• A new restraint type, IMPROPER dihedral, has been added. This restraint can be used in place of a ANGLE restraint in the case of nonplanar atoms with three ligands.  IMPROPER is better at maintaining the pyramidal shape of these groups than ANGLE.  Since the improper dihedrals can be calculated from the angles one should never attempt to restrain both at the same time.

• One can now PUNCH transformation matrices in several different coordinate systems.  Available options are TNT, PDB, GRID, and FRACTIONAL.  To do so add "SYSTEM PDB", for example, to the end of the PUNCH command line.

• Overlay will now limit B and OCC parameters of transformed models by the values given on the RANGE statement.

• This program was created to perform averaging over the ncs related portions of the asymmetric unit of a map.  One map is produced for each CLUSTER.  It does not loop over CLUSTER statements internally; you must give the program one PUNCH command for each map.  A prototype command file for running this program is given in tnt/doc/gather_map.

• When a map is calculated from Fourier coefficients the FOM is now used to calculate the "best" coefficient instead of blindly taking the F.  Since most maps are calculated using model phases which all have a FOM of one this change will not affect most operations.  It will allow the proper calculation of MIR maps.