On Jul 15, 2008, at 2:29 PM, Christoph Schmidt wrote:
> Dear all,
>
> this is kind of a follow up of the "cis proline cipp patch" email
> from last Friday.
You'd be better off asking it on the CNS users list, rather than the
xplor-NIH list, since we have separate topology and parameter files.
> QUESTION 2:
>
> This really puzzles me:
> There were 2 error messages asking for 2 IMPRopers which in its
> kind haven't been defined in the original parallhdg5_3.pro.
> A similar improper can be found in the topology-file (improper HA
> N C CB !defining the chirality of CA), but I could not find
> any defining values for this
> in the parameter file. (And when I calculated structures with all
> PRO set trans - just for comparison - there was no error message
> for these 2 IMPRopers meaning they are only used when calculation
> with the PRO set cis ???)
>
> So I had to "make them up" myself. Just in order to find out if I
> would be able to create a template pdb, I just put a random value
> with a very small "rating" value:
>
> IMPRoper HA NH1 CCIS CH2E 0.1000 {sd= 0.031} 0
> -72.1561
> IMPRoper HA N CCIS CH2E 0.1000 {sd= 0.031}
> 0 -72.1561
>
> Indeed, I got a template pdb and the subsequent structure
> calculation went "alright".
>
> DOES ANYONE HAVE THE RIGHT VALUES FOR THESE TWO IMPRopers OR
> WHAT IS THE BEST METHOD OF GETTING THEM?
You could take a look in the xplor-nih protein.top and protein.par
files for the equivalent impropers. In general, the way to get good
target values for impropers is to generate coordinates with pretty
good bond length and bond angle values, along with approximate
improper values. Then turn off the impropers and minimize the bonds
and angles again. Finally, read off the optimal improper values from
your coordinates.
In the classic xplor script, which ought to work for both CNS and
xplor-NIH, that might be something like
struct @whatever.psf end
param @initial.par end
vector do (x = random()) (all)
vector do (y = random()) (all)
vector do (z = random()) (all)
flags exclude * include bond angle impr end
mini powell nstep = 1000 end
flags exclude * include bond angle end
mini powell nstep = 1000 end
pick dihedral (sel 1) (sel 2) (sel 3) (sel 4) geom
stop
Assuming that the first minimization doesn't get stuck, and you set
up the selections in the pick dihedral command to match your improper
definition, you should be able to get optimized values of the
impropers, with which you could update your parameter file.
> (WHY IS THERE A NEED TO DEFINE 2 MORE IMPRopers IN THE cis-
> calculation WHEN THE trans-calculation DIDN'T NEED THEM?)
Perhaps the cis proline patch introduces a new chemical type?
Hope this helps.
--JK
