Vivek Sharma wrote: > I am having some troubles in putting dihedral restraints in the system I > am modelling.
> > So for e.g. I put a dihedral restraint (say phi for e.g.) for a residue > like this: > > SET ATOM_IDS = 'C:317' 'N:318' 'CA:318' 'C:318' > ADD_RESTRAINT RESTRAINT_PARAMETERS = 3 1 3 11 4 2 0 -1
0.261
This should be fine, provided that you really want the mean to be -1 radians, and that this is within the 'special_restraints' routine or similar:
http://salilab.org/modeller/manual/node32.html
Yes I am using 'special_restraints'.
> Now when I run modeller with such restraints, in the finally obtained > models, I dont see these exact values of restraints,
> which we mentioned explicitly, taken in by the model. Why this is so? > The values of these mentioned dihedral restraints in > the modelled models are far from the prescribed values.
You should check the log file to make sure the restraint is actually
there, and the end of the .rsr file.
Yes the restraints explicitly mentioned by me are there in the log file as well as in the *.rsr file.
> Now as I said above that I am giving 15deg as std. deviation,but when I > use smaller values like 5deg or so, the program is
> getting failed, it seems that it is getting difficult for it to optimise > the system, with smaller values of std. deviation. > > Last but not least, what I tried or trying is that creating the gaps in
> this section of 5 residues, where I am imposing > backbone dihedrals, so that MODELLER doesnt find any equivalent residues > in that position in templates and probably reduces > the number of restraints, and in that case I may be able to impose my
> restraints well enough for these five residues, but > still results doesnt seems to go hand in hand. Any comments in this > concern also will be helpful.
If you actually are applying the restraint, it is probably being
violated because it clashes with an existing restraint. Modeller will apply dihedral restraints to your system regardless of whether homology information is available (if unaligned, it will just use CHARMM parameters). Therefore you should remove the existing restraints first:
see http://salilab.org/modeller/FAQ.html#4 and the 'cispeptide' script in modlib/modeller/scripts/cispeptide.py (or the TOP equivalent in bin/__cispeptide.top) for an example.
Actually, here I didnt got the point completely. I did a little checking like
this, please see:
Template sequence : FINR
Target sequnece : FINR
PDB available for template. The target seq. has been modelled in various ways to check what kind of restraints are generated for a residue (N here, the third residue Asparagine), when there is a exact residue in the template at the equivalent position OR when there is a gap in the equivalent position OR when there is a different residue in the equivalent position. Now if we see below then the third residue has a different cases in the equivalent position in these three separate cases. I then checked restraint generated for each atom of the third residue
i.e. N in target.
Cases: FINR FI-NR FINR-
FINR FINR- -FINR
Results: The second case in which there is a gap in equivalent position in template i.e. for N, the restraints generate were least (for the atoms of this residue), clearly. And, the restraints are :
N stereo + phi-psi + omega + chi1
CA stereo + phi-psi + omega + chi1 + chi2
CB stereo + chi1 + chi2
CG stereo + chi1 + chi2
OD1 stereo + chi2
ND2 stereo
C stereo + phi-psi + omega
O stereo
I havent mentioned the restraints for the two other cases, but clearly the restraints (type and no.) were different from these. ""So I kind of consider it safe to create a gap at the position where I want to put my own restraints on, because in this case the MODELLER generate restraints are least"".
Now if you remember, then I had this problem that values which I was using for restraints was not imposed properly in the system modelled and you suggested that there might be some restraint clashes and if we see above case, in which there is gap and MODELLER generated restraints are least, we still have Phi-Psi binormal restraints. They probably might be clashing with the Phi-Psi values I am imposing (I am imposing Phi/psi dihedral angle restraints separately not the binormal ones), so I deleted them (Phi-psi binormal ones, which are generated by MODELLER and I checked log files and *.rsr that they are deleted) and generated the structure, but again no result, I still have restraint violations OR the values are very far from the values mentioned.
So I am not sure what I am missing or doing wrong, but it seems to me that probably some awkward values of dihedrals can not be modelled in MODELLER, though the values I am using to model are not awkward as such because they are derived from one protein (X-ray), I dont know the solution of this problem at the moment.
If you have any comments and they are very welcome. I am still trying few other options and checking, I will let you know whenever I have something in this concern. But any opinions will be great.