Re: soft-core FEP in namd 2.7b

From: Jerome Henin (jhenin_at_cmm.chem.upenn.edu)
Date: Sun May 10 2009 - 22:19:52 CDT

> @Jerome:
> ok, the lambda=1 case is fine with me :-)
>
> however, for the lambda=0 case:
>
>> Lennard-Jones (and of course it goes to zero as lambda vanishes). So
>
> This is right for lambda_up_LJ, but what about lambda_down_LJ at lambda=0?

Well, then lambda_down_LJ is equal to 1. Substitute in soft-core
potential, lather, rinse, repeat.

> a) is there simply no soft-core potential for the vanishing set of
> particles?

They feel exactly the same potential, sampled along the same
pathway... in the opposite direction.

> b) the vanishing set of particles are not expected to have clashes

Not when they *start* vanishing (lambda = 0), but they do when they
have almost completely vanished (lambda ~ 1).

> if a): @Chris: maybe one could stress explicitly what interactions (what
> with what) are considered for the soft-core potentials?

All the LJ interactions affected by the scaling, that is (say we are
mutating group A into B in environment E):
1) A <-> E
2) B <-> E

Plus, iff decouple is OFF:
3) A <-> A
4) B <-> B
(with decouple on, those interactions are not perturbed)

Side note: normally there are no A <-> B interactions.

> if b): is "fepVdwShiftCoeff=5.0" a little high, considering that even
> non-clashing particles have distances of 3A or so?

The shift applies to square distances and is given in A^2. The default
shifting distance is really sqrt(5) in Angstrom.

> Sorry, I lost the link to your original soft-core paper, if you could send
> it to me again?

There (not ours, by the way):
http://link.aip.org/link/?JCPSA6/100/9025/1

Cheers,
Jerome

> Jerome Henin wrote:
>>
>> Hi Sebastian,
>>
>> There is no need to neglect anything, the soft-core formalism is
>> analytically accurate. Have a look at the expression for the soft-core
>> potential and let lambda be 1: then it becomes identical to a 6-12
>> Lennard-Jones (and of course it goes to zero as lambda vanishes). So
>> switching from L-J to soft-core only changes the intermediate states 0
>> < lambda < 1, not the end-points of the calculation, hence it does not
>> change the (converged) free energy difference.
>>
>> Best,
>> Jerome
>>
>> On Sun, May 10, 2009 at 11:36 AM, Sebastian Stolzenberg
>> <s.stolzenberg_at_gmail.com> wrote:
>>
>>>
>>> Thank you, Chris, I missed that link,
>>>
>>> theres is one more thing attracting my curiosity, the vdW
>>> hard-core<->soft-core transitions:
>>>
>>> The real vdW potential is hard-core, but NAMD2.7b FEP can use soft-core.
>>> Can one really neglect \delta_G from a hard-core<->soft-core transition?
>>> How could one measure \delta_G and thus convince oneself in practice?
>>>
>>> Thank you,
>>> Best,
>>> Sebastian
>>>
>>>
>>> Chris Harrison wrote:
>>>
>>>>
>>>> Sebastion,
>>>>
>>>> Please look here to begin:
>>>> http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l/9953.html
>>>>
>>>>
>>>>
>>>> C.
>>>>
>>>>
>>>> --
>>>> Chris Harrison, Ph.D.
>>>> Theoretical and Computational Biophysics Group
>>>> NIH Resource for Macromolecular Modeling and Bioinformatics
>>>> Beckman Institute for Advanced Science and Technology
>>>> University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
>>>>
>>>> char_at_ks.uiuc.edu <mailto:char_at_ks.uiuc.edu>
>>>>  Voice: 217-244-1733
>>>> http://www.ks.uiuc.edu/~char <http://www.ks.uiuc.edu/%7Echar>
>>>>  Fax: 217-244-6078
>>>>
>>>>
>>>>
>>>> On Fri, May 8, 2009 at 11:57 AM, Sebastian Stolzenberg
>>>> <s.stolzenberg_at_gmail.com <mailto:s.stolzenberg_at_gmail.com>> wrote:
>>>>
>>>>   Dear All,
>>>>
>>>>   I have some trouble interpreting the new parameters (listed below)
>>>>   in the namd 2.7b manual's FEP chapter.
>>>>   Here are my interpretations, I would be glad if you could check
>>>>   them with me:
>>>>
>>>>   I assume that decouple is set to "on".
>>>>
>>>>   manual p. 118:
>>>>   "fepVdwShiftCoeff / tiVdwShiftCoeff":
>>>>   considered are only vdW interactions between the growing/shrinking
>>>>   particles and their respective environments only.
>>>>   (otherwise: what's the use of soft-core for vdW of the environment
>>>>   with itself?)
>>>>
>>>>   "fepElecLambdaStart/tiElecLambdaStart ":
>>>>   considered are only the elect. interactions of the growing
>>>>   particles with its environment:
>>>>   E_el(0.5)=0 linearly increased to E_el(1.0)=full strength
>>>>   (my thinking is that interactions between the shrinking particles
>>>>   with the environment are *decreasing*)
>>>>
>>>>   p. 119:
>>>>   "fepVdwLambdaEnd / tiVdwLambdaEnd":
>>>>   it means that @lambda>0.5, we set fepVdwShiftCoeff
>>>>   (tiVdwShiftCoeff) to zero.
>>>>
>>>>   Is this all correct?
>>>>   If yes, then @lambda=0, I see an abrupt transition from
>>>>   "hard-core" to "soft-core" for vdW interactions between shrinking
>>>>   particles and their environment.
>>>>   Is the corresponding free energy difference negligible with the
>>>>   default fepVdwShiftCoeff value of 5A^2? If not, how can one
>>>>   practically measure this free energy difference?
>>>>
>>>>   Thank you,
>>>>   Best,
>>>>   Sebastian
>>>>
>>>>
>>>>
>>>
>>>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:46 CST