Department of Chemistry and Chemical Biology
Questions and problems?

Constraints and Restraints

Nomenclature: the terms `constraints' and `restraints' are often used interchangeably. Constraints implies absolutely fixed values - the only ways to achieve these generally within AMBER is to use the `belly' option (Cartesian coordinates) or SHAKE (bond lengths). [Gibbs also allows bond, angle and dihedral constraints - see INTR in the Gibbs manual or read the Interface/AMBER manual.]

Restraints implies use of an energy function without absolute fixing of the desired quantity; the refc file can be used for restraining Cartesian coordinates of selected atoms, or internal coordinate restraints can be applied in Parm (not recommended) or Sander (more flexible).

... how to constrain a dihedral

There are two methods, one on card 6 of the Parm input [harmonic] and a second in Section THREE of Sander input [flat-well/parabolic/linear hybrid]. There is no way to keep torsion angles absolutely fixed within AMBER [except in Gibbs - see above]. Incidentally, you can easily create a energy vs. torsion table using Interface.

Is it possible to operate two independent sets of distance angle restraints, allowing both to be initially ramped in and then after 20-25ps the [e.g.] Watson/Crick distance and angle restraints to be slowly removed leaving the NMR derived restraints in place ??

Dave Case:

Yes. The variable IFVARI can be used to control individual restraint weights. See p. 97 of the manual. For example, if you wanted to have a WC constraint on for the first 2500 steps, then ramp to zero during the next 2500 steps, you might do something like the following: 

 &rst  iat=?,?, r1=???, r2=???, r3=???, r4=???, rk2=???, rk3=???,
     nstep1=0, nstep2=2500,   &end
 &rst  iat=?,?,  nstep1=2501, nstep2=5000, ifvari=1,
      r1a=???, r2a=???, r3a=???, r4a=???, rk2a=0.0, rk3a=0.0,   &end
The first &rst line sets up the constraint for the first 2500 steps, and it is constant (ifvari has the default value of zero.) The second &rst continues this constraint for another 2500 steps, but ramps the force constants rk2 and rk3 down to zero; as these get smaller, there essentially becomes no penalty for violating the constraint. After 5000 steps, neither constraint is active.
How can we make the file 'refc' for running with internal constraints?

Refc cannot be used for internal (bond, angle or dihedral) constraints: it is for Cartesian restraints only. The format is the same as inpcrd and restrt.

I would like to ask you about distance restraints in Sander. I looked at /amber41/demo/plastocyanin/ to get hints for making input files. I could not understand which of r1,r2,r3,and r4 apply to the longest distance and the shortest distance of two protons. The document of Sander indicates the need for distance when E=0, but I have no such data. I would like to use only two data mentioned above as restraints. Is it possible to omit one of r1,r2, r3,and r4? And which of them must be defined? Furthermore, I would like to know how to decide RK2 and RK3.

R1, R2, R3, R4 define a flat-welled parabola which become linear beyond a specified distance. I.e. 

         \                       /
          \                     /
           \                   /
            .                 .
              .             .
                 ._______.

           R1    R2      R3   R4

      "\" = lower bound linear response region 
      "/" = lower bound linear response region 
      "." = parobola
      "_" = flat region


If you have determined lower and upper bounds from an NMR experiment,
those would typically correspond to R2 and R3. Note that the flat well
means that any value R2 <= value <= R3 is equally acceptable. R1 and
R4 define linear response regions. These are sometimes used so that 
restraints that severely violate the lower and upper bounds don't
tear the structure up. A typical value of R1 is R2-2.0. And a typical
value of R4 is R3+2.0 (angstroms).



What is the form of the potential used for chemical shift restraints?



Answer is in the manual, but somewhat hidden!   See the discussion
about the "IPNLTY" variable:

      IPNLTY
                 = 1   the program will minimize the sum  of  the
                       absolute  values  of  the  errors; this is
                       akin to minimizing the crystallographic R-
                       factor (default).

                 = 2   the  program will optimize the sum of the
                       squares of the errors.

                 = 3   For NOESY intensities, the penalty will be
                       of the form

                           awt [Ic^(1/6)-Io^(1/6)]^2.

                       Chemical  shift  penalties  will be as for
                       ipnlty=1.




Further discussion appears in section SIX, "Chemical Shift Restraints".


Basically, the program will minimize the sum of the absolute values of
the "errors", where in this case the "error" is the difference between
the calculated and the observed shift.  The SHRANG variable allows you
to ignore errors less than some cutoff, and the WT variable allows you
to weight some shifts more heavily than others in the sum.



How about using positional restraints to force conformational changes?



Positional restraints  work well enough to keep a molecule near 
its starting conformation, and they can force a molecule from one 
conformation to a second SIMILAR conformation.  But they don't do 
a good job of forcing a molecule through substantial
regions of conformational or distance space.





Web Masters <webadmin@www.amber.ucsf.edu>


Last modified: Mon Sep 18 14:15:42 PDT 1995

Department of Chemistry and Chemical Biology, 102 Hurtig Hall, 360 Huntington Ave., Boston, MA 02115
Tel: 617-373-2822 Fax: 617-373-8795

Send mail to e.chudin@neu.edu with questions or comments about this web site.
Last modified: Tuesday, April 01, 2003