Re: [AMBER] TI with softcore at lambda=0 or 1?

From: <steinbrt.rci.rutgers.edu>
Date: Thu, 9 Aug 2012 11:08:40 -0400 (EDT)

Hi,

check e.g.

Title: Soft-Core Potentials in Thermodynamic Integration: Comparing One-
and Two-Step Transformations
Author(s): Steinbrecher, Thomas; Joung, InSuk; Case, David A.
Source: JOURNAL OF COMPUTATIONAL CHEMISTRY Volume: 32 Issue: 15
Pages: 3253-3263 DOI: 10.1002/jcc.21909 Published: NOV 30 2011

for the effect of scbeta. The default value should be ok for most
situations, but is certainly one variable you could optimize. There is no
correct or incorrect value for scalpha and scbeta, but they influence the
efficiency of your sampling in a case-dependent manner.

Thomas

On Thu, August 9, 2012 9:37 am, Liu Denis wrote:
> Dear Dr. Thomas:
> If I want to use one-step soft core, should I explicitly set the
> scbeta parameter in input file?
> I did not set it explicitly because there is a default value in Amber
> Manual. May it be the reason that I got a strange curve?
>
> Best Regards!
> Denis
>
> 2012/8/9 <steinbrt.rci.rutgers.edu>:
>> Hi,
>>
>>> the dvdl curve near lambda 0 or 1 seems to be rather steep, far from
>>> linear. Is it normal?
>>
>> that looks unexpected to me. The appearing/disappearing vdW potential is
>> quite flat close to the endpoints, so it should by definition not give
>> large dvdl values. It is hard to say without knowing your system and
>> setup. Do you have vdW-radii changing that are not in the softcore mask?
>> Or big changes in solvated charges? Your curve may be perfectly fine,
>> but
>> a closer look at what goes on in your system seems warranted...
>>
>> Thomas
>>
>>>
>>> What reason may cause such result?
>>>
>>> Best Regards!
>>> Denis
>>> 2012/7/26 <steinbrt.rci.rutgers.edu>
>>>>
>>>> Hi,
>>>>
>>>> > The post http://archive.ambermd.org/201002/0108.html </>said
>>>> softcore
>>>> > cannot use at lambda <0.01 or >0.99,
>>>> > Does it support now?
>>>>
>>>> no, that limitation has not changed. There is some work in progress to
>>>> make a new version of TI that would allow l=0/1, but it is not
>>>> finished
>>>> yet. However, I have never seen this as really problematic. The dvdl
>>>> curve
>>>> should be linear at both ends and you really want the integral over
>>>> it,
>>>> so
>>>> using 0.01 or 0.00 as a fixed-point for numerical integration amounts
>>>> to
>>>> the same thing. Unless you want more than a hundred lambda-windows,
>>>> there
>>>> is no real gain from simulating in the <0.01 / >0.99 region.
>>>>
>>>> > If I want to get the point at Lambda=0, do I have to use "hardcore"?
>>>>
>>>> if you really want this point, then yes, you can not use soft core
>>>> potentials.
>>>>
>>>> > Can I mix the result using hardcore at 0 or 1 and using softcore at
>>>> other
>>>> > lambda? In other words, at lambda=0 or 1, does softcore and
>>>> hardcore
>>>> give
>>>> > out the same result theoretically?
>>>>
>>>> No! Far from it actually, the two approaches amount to very different
>>>> potential functions and should not even be close in terms of curve
>>>> shape.
>>>>
>>>> Kind Regards,
>>>>
>>>> Thomas
>>>>
>>>> Dr. Thomas Steinbrecher
>>>> formerly at the
>>>> BioMaps Institute
>>>> Rutgers University
>>>> 610 Taylor Rd.
>>>> Piscataway, NJ 08854
>>>>
>>>> _______________________________________________
>>>> AMBER mailing list
>>>> AMBER.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>>
>>>
>>>
>>> --
>>> ----GnuPG Public Key----
>>> Key ID: 0x488C9571
>>> Fingerprint: C292 D888 D713 4D13 57F0 C07D 16F3 10C9 488C 9571
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>
>>
>> Dr. Thomas Steinbrecher
>> formerly at the
>> BioMaps Institute
>> Rutgers University
>> 610 Taylor Rd.
>> Piscataway, NJ 08854
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
>
>
> --
> ----GnuPG Public Key----
> Key ID: 0x488C9571
> Fingerprint: C292 D888 D713 4D13 57F0 C07D 16F3 10C9 488C 9571
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>


Dr. Thomas Steinbrecher
formerly at the
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 09 2012 - 08:30:03 PDT
Custom Search