Re: RATTLE failure

From: Himanshu Khandelia (hkhandel_at_memphys.sdu.dk)
Date: Sun Sep 09 2007 - 08:35:01 CDT

Philip,

In my personal experience, there are problems with the way SHAKE has been
implemented in NAMD. When I use SHAKE and use a 2 fs timestep, there is
significant drift in the simulation.

In CHARMM, shake is well implemented, but I would recommend you to stay
away from it in NAMD, especially if you have an anisotropic system like a
lipid bilayer. Stick to a 1fs time step, and use 2 and 4 for nonbondedfreq
and fullelectfrequency.

Hope this helps,

-himanshu

----------------------------
Himanshu Khandelia, PhD
Research Assistant Professor (Postdoc)
MEMPHYS, Center for Membrane Physics: www.memphys.sdu.dk
University of Southern Denmark (SDU)
Campusvej 55,
Odense M 5230, Denmark

Phone: +45 6550 3510
Fax: +45 6550 4048
email: hkhandel_at_memphys.sdu.dk
WWW: www.memphys.sdu.dk/~hkhandel
-----------------------------

On Sat, 8 Sep 2007, Philip Peartree wrote:

> I will try letting the simulation run free. In production I would do, so I
> don't
> see the harm. I did do minimisation before thermalising the enzyme. I believe
> RATTLE is separate to the algorithm that fixes the atoms. I just wondered
> whether the fixed nature of the atom might be causing the RATTLE algorithm (I
> think I have settle on) to fail. Only one way to find out!!!
>
>
> Philip Peartree
>
>
> Quoting Eric Perim <eric.perim_at_gmail.com>:
>
> > As far as I know constraining the backbone will only reduce de computational
> > cost of the simulation, since it will reduce the degrees of freedom, but I
> > guess a totally free equilibration will be at least as effective as this
> > one, but surely more time consuming. If the structure was not subject to any
> > kind of geometry optimization before the equilibration I guess it would be
> > good to let it free.
> > One important thing is that the RATTLE algorithm is used to keep the bonds
> > rigid, I don't know if it is the same as the one used to keep the atoms
> > fixed. If it is a different algorithm, then it may be worth a try to fix the
> > atoms through the fixedatoms options, instead of making the bonds rigid.
> >
> > Eric
> >
> >
> > On 07/09/2007, Philip Peartree <P.Peartree_at_postgrad.manchester.ac.uk> wrote:
> > >
> > > Thanks Eric, I had contemplated it! One thing I am wondering, atom
> > > referenced in
> > > the error is a backbond carbon (incidentally part of a glycine) and at
> > > the time
> > > I am running with fixed backbone.
> > >
> > > When I used to run in charmm (with the same sim) I used to constrain the
> > > backbone when equilibrating and then release the backbone. Is there any
> > > point
> > > in either of these things or would a totally free equilibration be better?
> > > Up
> > > to this point I've heated to 300K (totally free) and equilibrated the
> > > waterbox
> > > (with the protein fixed)
> > >
> > > I reckon this might solve the error, as I have seen in other posts that
> > > fixing
> > > atoms introduces unreasonable forces into the system.
> > >
> > > Philip Peartree
> > >
> > > Quoting Eric Perim <eric.perim_at_gmail.com>:
> > >
> > > > Hi Philip,
> > > >
> > > > I got this error more than once, what I did to solve it was turn off the
> > > > rigibonds option (left only the solvent rigid). It may not be the best
> > > > solution but it works.
> > > >
> > > > Eric
> > > >
> > > >
> > > > On 04/09/2007, Philip Peartree <P.Peartree_at_postgrad.manchester.ac.uk>
> > > wrote:
> > > > >
> > > > > Hi All,
> > > > >
> > > > > I'm getting a "ERROR: Constraint failure in RATTLE algorithm" in my
> > > > > simulations,
> > > > > anyone know what could cause this? I read on the list that increasing
> > > the
> > > > > Margin parameter could help, so I set it at 1.0, would further
> > > increases
> > > > > help
> > > > > or is it something more fundamental? My sim is a protein in an
> > > > > 80x70x80
> > > > > water
> > > > > box, with PME and this part of the simulation it fails in has the
> > > backbone
> > > > > atoms fixed but all others free? It happens after about 200ps.
> > > Previously
> > > > > (before setting the Margin parameter) it happened after about 30ps.
> > > > >
> > > > > Philip Peartree
> > > > >
> > > > >
> > > >
> > >
> > >
> > >
> > >
> >
>
>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:46:57 CST