AMBER Archive (2009)

Subject: Re: [AMBER] Question of about MM-PBSA and mm_pbsa_createinput.pm script

From: David A. Case (case_at_biomaps.rutgers.edu)
Date: Tue Jun 09 2009 - 08:20:22 CDT

On Tue, Jun 09, 2009, Samuel Genheden wrote:
>
> I had trouble with the igb=5 and igb=7 options in MM/GBSA calculations
> and after some time I found a strange thing in the
> mm_pbsa_createinput.pm script.
> This is the following code-snippet, from the create_sander_input subroutine:
>
> if($igb == 1 || $igb == 2){
> $offset = 0.09;
> }
> else{
> $offset = 0.00;
> }
>
> Why should the offset be any different between, e.g. igb=2 and igb=5?
> I cannot find any explanation either in the Amber 10 manual or the
> original articles of igb=2/5 or igb=7. In my opinion, this is bug.

I agree. The list is looking for a good volunteer to take over
maintenance of the mm-pbsa scripts. My understanding is that new ideas
are in development at Florida(?), but I haven't seen much yet along these
lines.

My $.02: MM-PBSA is actually a fairly complex calculation (especially with the
gazillion options that have been put into the perl scripts), and current
scripts are not well suited for beginning users. Everyone using this
technique should first do some simple calculations by hand, then study
carefully what the perl scripts are doing. Users should expect to need to
(occasionally?) track down problems by doing the individual steps themselves.

To remedy this: (a) improve the documentation, test cases and tutorials; (b)
work much harder on error identification in the scripts; (c) make the output
easer to understand.

Even with all this, interpreting and using mm-pbsa output is partly an art.
The model assumes that one has stable trajectories that are sampling the right
regions of configuration space. If the system is still drifting, or if
conformational changes happen during simulations, one can get completely
unreliable results.

....dac

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber