Hi, Thomas
and thanks for your quick answer. When I read the manual ifsc=2
sounded ideally for my case, but as you explain it I would have to use
ifsc=1. A good input option, in my case, would be to have a option
similar to scmask that will zero the Lennard-Jones intermolecular
interactions with the masked atoms. Now I have to prepare another
prmtop file where I manually set the parameters to zero.
/ Samuel
2010/1/13 <steinbrt.rci.rutgers.edu>:
> Hi Samuel,
>
>> (nmropt=1) to the benzene molecule. To perturb to nothing I use the
>> option ifsc=2, as explained in the manual. But when I run sander it
>> stucks at a particular place in the initialization phase. I have tried
>
> I can't comment much on the energy decomposition you want to use, but to
> conduct the TI calculation you describe, you would not use ifsc=2. This is
> a special case that only concerns perturbing the complete system to
> nothing, which can then no longer run in parallel, since you have no
> second prmtop file. A protein+benzene -> protein perturbation would still
> run with ifsc=1.
>
> ifsc=2 implies a serial run, this may very well be the cause auf the
> hangup you see, since TI decomp assumes you run a multisander-job with two
> groups.
>
>> to set ifsc to 0, and then sander runs perfectly fine. By looking
>
> It's suprising that you can run this with ifsc=0, because your two systems
> would have different numbers of atoms and that would cause a TI
> calculation to crash without the soft core options to handle extra atoms.
>
> Looking at your input file:
>
>> icfe=0, clambda=0.90,
>
> This would not be a TI run at all if you set ifsc=0. You have to set
> icfe=1 if you want to use TI, only in the special case of ifsc=2, which
> doesnt apply here, can you run a TI-like calculation without icfe>0.
>
> I guess the documentation could be clearer about ifsc=2, which has become
> more of a legacy option, since the internal energies are now not scaled by
> lambda, so perturbing a complete system to nothing should always give you
> dVdl=0. The option may even disappear from future versions of Amber...
>
> Kind Regards,
>
> Thomas
>
> Dr. Thomas Steinbrecher
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jan 13 2010 - 03:00:04 PST
