[gmx-users] free energy perturbations and B-parameter specifications

[Erik Lindahl]

On Jul 17, 2007, at 4:45 PM, Soren Enemark wrote:

> Dear Gromacsers,
>
> I have been doing free energy perturbations Gly->Ala in one strand  
> of collagene.
>
> Now, the problem is that I have later closely reread the section  
> 5.6.4 ("Free
> Energy Calculations") p. 105 in which it says:
> "Bonded interactions between atoms that are not perturbed do not  
> need B parameter specifications ..".
>
> My questions are:
> 1. Does this mean that bonded interactions between atoms that ARE  
> perturbed DO need B parameter specifications? Even if they are the  
> same in state
> A and B?
> 2. What about the other topology parameters (angles, dihedrals  
> etc), do they need to be specified too?
> 3. What happens if I have not specified these specifications?

Since a couple of years we automatically copy the A state parameters  
to the B state if the latter aren't specified.

So, there are really three options e.g. for an entry in the [ bonds ]  
section:

1. You only specify the bond atoms & type without any parameters  
(most common approach)

      In this case, the A state parameters will be assigned based on  
the atomtypes in the A state, and the B state parameters likewise  
based on the B state atomtypes.
      The interaction is perturbed.

2. You provide the A state parameters explicitly, but not any B state  
parameters:

     Since you've decided to override the automatic assignment your  
choice will be used in both the A and B states (i.e., the interaction  
will not be perturbed)


3. You provide both A and B state parameters explicitly:

      Your parameters will be assigned to each state, and the  
interaction is perturbed.



Cheers,

Erik