Re: [ccp4bb] CNS v1.2 and unwanted introduction of OXT
On Mon, 5 Mar 2007, James Irving wrote:
Hi Fred,
From memory, this can occur due to long bond lengths in the model being
interpreted as chain breaks, try editing the generate.inp or
generate_easy.inp script and increase the value for break_cutoff:
{* cutoff distance in Angstroms for identification of breaks *}
{* the default of 2.5A should be reasonable for most cases. If the input
structure has bad geometry it may be necessary to increase this
distance *}
{===} break_cutoff=2.5;
You may also try switching off the automatic detection of mainchain breaks:
* automatically detect mainchain breaks in proteins based on distance *}
{* the peptide link at break points will be removed *}
{+ choice: true false +}
{===} auto_break=true;
It is also worth checking that the offending residues have their full
complement of backbone atoms in appropriate places (not accidentally
zapped to far-flung regions of coordinate space).
Cheers,
James
Hi James,
I tried to increase the parameter value to 3.0 A. The resulting file gives
the same behaviour (and the newly introduced OXTs are within 0.05 A of the
N atom of the following residue) so I think the distances are reasonable.
So unfortunately I think your valuable suggestion does not solve the
problem. On the graphics with Coot the starting model looks fine in these
regions.
I need to keep the automatic mainchain break detection because there are
some gaps in each chain of the model (12 chains in all).
So if there are other suggestions to solve this br of a problem, they
are most welcome!
Fred.
--
Fred. Vellieux, esq.
=
IBS J.-P. Ebel CEA CNRS UJF / LBM
41 rue Jules Horowitz
38027 Grenoble Cedex 01
France
Tel: (+33) (0) 438789605
Fax: (+33) (0) 438785494
=
Re: [ccp4bb] CNS v1.2 and unwanted introduction of OXT
On Mon, 5 Mar 2007, Fred. Vellieux wrote: Hi James, I tried to increase the parameter value to 3.0 A. The resulting file gives the same behaviour (and the newly introduced OXTs are within 0.05 A of the N atom of the following residue) so I think the distances are reasonable. So unfortunately I think your valuable suggestion does not solve the problem. On the graphics with Coot the starting model looks fine in these regions. I need to keep the automatic mainchain break detection because there are some gaps in each chain of the model (12 chains in all). So if there are other suggestions to solve this br of a problem, they are most welcome! Fred. Hello again, I believe that I found the cause of the problem and a solution to it: not by looking at the molecule with Coot but by looking at the separate PDB input files where the problems occur. They all share the same features: the chain label used by CNS is kept along within Coot. For the parts that are modelled de novo using Coot, the output pdb file does not contain the CNS chain label. It is only at those positions where there is a discontinuity for chain label present or absent does CNS introduce an extra OXT atom. I have introduced the missing chain labels where necessary, and I am testing this right now: the file produced by the generate script now does not contain added OXT atoms. I thought I'd inform the bulletin boards of this annoying feature. Fred. -- Fred. Vellieux, esq. = IBS J.-P. Ebel CEA CNRS UJF / LBM 41 rue Jules Horowitz 38027 Grenoble Cedex 01 France Tel: (+33) (0) 438789605 Fax: (+33) (0) 438785494 =
