Date: Fri, 12 Jul 2002 11:58:02 -0700
From: "David A. Case"
Subject: are there inconsistencies in anal?


On Fri, Jul 12, 2002, Prabal Maiti wrote:
> I did the analysis for the same 10-mer DNA as in my previous email with
> anal in AMBER6.
>
> Here also I see that the group energy are different at the begining and at
> the end (in the order they are printed in anal output).
>
> ENERGY CONTRIBUTION BY GROUPS
> 1 -149.472000
> 2 -196.461000
> 3 -120.201000
> 4 -169.040000
> 5 -159.504000
> 6 -191.575000
> 7 -168.044000
> 8 -64.154000
> 9 -125.126000
> 10 -114.397000
> 11 -132.681000
> 12 -166.969000
> 13 -67.345000
> 14 -139.577000
> 15 -148.896000
> 16 -151.668000
> 17 -157.445000
> 18 -49.034000
> 19 -119.173000
> 20 -110.527000
> 21 -16819.504000
>
>
>
> INDIVIDUAL GROUPS INTERACTION
>
> 1 -176.330000
> 2 -274.440000
> 3 -196.800000
> 4 -252.220000
> 5 -237.910000
> 6 -255.210000
> 7 -217.450000
> 8 -117.980000
> 9 -170.870000
> 10 -146.200000
> 11 -146.080000
> 12 -210.150000
> 13 -93.820000
> 14 -200.570000
> 15 -218.440000
> 16 -173.840000
> 17 -190.140000
> 18 -95.250000
> 19 -165.000000
> 20 -145.290000
> 21 -20043.610000
>
>
> Now which group energy is correct.

Both are "correct" but different. The "energy contribution by groups" takes
the "intra-group" terms and adds half of the the "inter-group" terms.
In this way, the sum of the individual groups gives the total energy.
This is the "ordinary" method of energy decomposition.

The "individual groups interaction" assigns the "inter-group" energy to
both partners, so that the total terms don't add up to the total energy
(since the inter-group parts are double-counted.) I think the authors
mainly expected people to analyze the interaction matrices. I would agree
that the total values printed are somewhat un-intuitive (but there is
probably someone out there for whom this is exactly what she wants!)

The easiest way to see what is going on is probably to study the output from
an example where there are only two groups: e.g.
$AMBERHOME/test/vac_rna/analout.save.

> For PME/EWALD I get the effective field acting on each atom and and
> multiplied by charge give me the energy for that atom. Sum of all the
> energy ofcourse gives the total energy
>

Note that anal does not know about PME. Your ideas for modifying sander
to do that are in the right direction, but there are things to watch out
for. In particular, the "PME field" would include contributions from the
1-2 and 1-3 interactions, and these would have to be separately removed.
And you would have to do N PME calculations (where N is the number of groups),
with charges only on the atoms within the group in question.

You might want to study the "decomp.f" routine in sander for some ideas
about doing anaylsis there (no PME, but some other aspects are provided,
as a part of the decomposition analysis done by MM-PBSA.

..hope this helps....dac