Re: [AMBER] MMPBSA Error: PB Bomb in pb_atmlist(): maxnba too short

From: Jason Swails <>
Date: Mon, 27 Oct 2014 08:19:41 -0400

On Mon, 2014-10-27 at 12:57 +0530, MOHD HOMAIDUR RAHMAN wrote:
> Dear Amber users and Developers
> I want to calculate per residue decomposition energy, for this I am using
> module along with Amber 12. I have six systems of varying
> concentration of solute. The three system which have low concentration
> (also less no of atoms) run successfully, while others three systems which
> have higher concentration (high atoms numbers) I am observing following
> error at the last of _MMPBSA_complex_pb.mdout.0 file .
> --------------------------------------------------------------------------------
> --------------------------------------------------------------------------------
> Cpptraj Generated Restart
> begin time read from input coords = 1.000 ps
> Number of triangulated 3-point waters found: 0
> --------------------------------------------------------------------------------
> --------------------------------------------------------------------------------
> Cpptraj Generated trajectory
> minimizing coord set # 1
> PB Bomb in pb_atmlist(): maxnba too short
> I search this error in Amber mailing list, but I didn't found any useful
> information. I given the following input file
> &general
> startframe=1001, endframe=2500, interval=50,
> keep_files=0,
> /
> &pb
> istrng=0.0,
> /
> &decomp
> idecomp=4, dec_verbose=0,
> /
> Would any ones suggest me to resolves this problem.

Your easiest bet would probably be to use GB instead for your
decomposition analysis. It is a bit more straightforward (even if it is
still not strictly correct) to decompose GB into pairwise interactions.
Not to mention, the GB calculation should be much faster (at least in my

To anticipate your argument that "PB is more accurate" -- well that's
true. GB models are often trained on PB results, and PB is more
theoretically rigorous than GB is. However, neither PB nor GB are
pairwise decomposable because the dielectric boundary (aka the effective
radii in GB) is an inherently non-local property. So what you are
computing is a fictitious quantity whose actual meaning is poorly
defined. Pairwise decomposition can be qualitatively useful to identify
trends -- but that is something for which GB should suffice.

If you really want to get PB pairwise decomp working, you can modify the
source code (look for the pb_atmlist subroutine) and increase maxnba
until the error goes away. At least that's what I would do, although I
can't guarantee it will work.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Mon Oct 27 2014 - 05:30:01 PDT
Custom Search