[gmx-users] TI/FEP, BAR, and topologies

[Michael Brunsteiner]

Hi,

I'd like to perform TI calculations as described in section 3.12.2 of the (version 4.5.4) manual.
my questions are:

1) i understand in Gromacs TI/FEP is implemented as a single topology, and not dual topolgy
algorithm, is that correct?


2)  to successfully analyze the results with BAR what is a good frequency at which i would

save energies and dU/dlambda values?
3) how do i set up topologies in practice?

If for example i wanted to mutate an ASN to a PHE residue in solution I seem to have at least two options:

  a) a direct mutation, where i'd mutate, e.g., CD1  and CD2 in PHE to OD1 and ND2 respectively in ASN,

and so forth, including a transformation of a few PHE atoms to dummies, and a transformation
of some bonded energy terms.


  b) construct a bastard residue from a PHE and a ASN residue, where the two residues share

the backbone atoms, including C-alpha, but where there are two full and different copies of the side chain -
and then mutate from state A with all sidechain atoms of, say, ASN set to dummies and the PHE atoms
fully turned on, to state B (vice versa) ... this would be like emulating a of dual topology approach,
wouldn't it?


is there any good reason for using either a or b? To me option a seems to be the more tedious one
as i'd have to include the changes in bonded energy terms by hand into the topology file, while with
option b pdb2gmx and grompp would do most of that work for me, correct? Also if i used option a
then at the ASN state would I not get some spurious effects from the ring topology of PHE, from the PHE
atoms that have their non-bonded interactions turned off but their bonded interactions are still there??

(i seem to recall that totally turning off any bonded interactions opens a can of worms)
thanks in advance for any help!

cheers
Michael