On Fri, Nov 19, 2004 at 03:50:31AM -0500, Youbin Tu wrote: > I want to define some restraints in the top file. Assume we know the > maximum and minimum distances between two atoms (non-bond), Which form > and group type should we use? since there are so many types of groups > to define distance restraints, which one should I use for my case? I > looked through the manual, but did not find enough talking about it. > > Here is an example, > The maximum distance between Fe:513 and C7:514 is > 6.0(1.0) > The miminum distance between Fe:513 and C7:514 is > 2.0(1.0) > I included the following line in the top file: > > SUBROUTINE ROUTINE = 'special_restraints' > ADD_RESTRAINT ATOM_IDS = 'FE:513' 'C7:514', ; > RESTRAINT_PARAMETERS = 2 1 1 22 2 2 0 6.0 1.0 > > ADD_RESTRAINT ATOM_IDS = 'FE:513' 'C7:514', ; > RESTRAINT_PARAMETERS = 1 1 1 21 2 2 0 2.0 1.0

This depends on what your PDF looks like. The way you've coded it, the distance will be found at 2.0 or 6.0 with maximum probability, which is probably not what you want. More likely you have a Gaussian distribution (form 3) centered around 4.0, which tails off to negligible probability above 6.0 or below 2.0.

If you have a histogram of probabilities as a function of distance, you could use form 10 (cubic spline).

> Again, If I did so, does that mean that the program have to force > the > distance between those 2 atoms within that range even though it may > take the risk to corrupt the whole structure? Or if some restraints > will destroy the whole structure, the program will discard the user's > restraints in order to " save" the final model?

In the case of conflicts between homology and user restraints like this, Modeller will try to fulfil both, and will probably end up with something which violates both. You should either remove any homology-derived restraints which conflict with your own, or decrease the standard deviation of your restraints to make them 'stronger'.

Ben Webb, Modeller Caretaker