RE: AMBER: molsurf error in MM/PBSA calculation

From: Paul Beroza <>
Date: Thu, 14 Jul 2005 16:01:08 -0700

You've found a bug in molsurf.

Here's what's happening:

The centers of the NH1, NH2, 2HH1, and 2HH2 atoms
in the arginine guanidinium group
form a rectangle (side lengths 1.007 and 2.309),
which is an unfortunate coincidence.

A "probe position" is the location of the
center of the spherical probe used to define
the molecular surface. A possible probe position
is identified as a point that is the proper distance
(probe_radius + atom_vdw_radius) from three atoms.

For each of these possible probe positions, other atoms are checked
to see if the probe overlaps with the van der Waals
surface of a 4th atom. If it does, the point is discarded.
If it does not, it is a "probe position".

In the case of the 4 arginine atoms, 3 atoms define
the point, and the 4th atom is tangent to the possible probe position.
Whether there is overlap is subject to floating point precision (i.e.,
random). The fact that the probe radius is 0 doesn't matter - in that
case the possible probe position is the point of intersection
of 3 of the atomic van der Waals radii, and the overlap check
is made between that point and the van der Waals' radius of
the 4th atom.

If you try the calculation with different probe radii,
every once in awhile you'll get a valid surface area
(i.e., the rounding just happened to result in the
proper identification of probe positions), but most of
the time, you'll get an error.

A workaround is to increase the radius of one of the
atoms by a little bit. Lame, but it works.

I'm surprised this bug hasn't surfaced before - I guess
you just got lucky.

I'll see what I can do about a fix, but it's been awhile
since I looked at this code.


-----Original Message-----
From: David A. Case []
Sent: Tuesday, July 12, 2005 6:24 PM
Cc: Paul Beroza
Subject: Re: AMBER: molsurf error in MM/PBSA calculation

On Wed, Jul 13, 2005, JunJun Liu wrote:
> Thank you for response! Attached please find the gzipped pqr file. The
> command I executed and the corresponding output are as follows:
> ---------
> $ molsurf test.pqr 0.0
> 5069 atoms read
> odd number of probe positions on torus!
> ---------
> Please help me resolve this problem, thanks in advance!

I have no clue myself...I reduced this to a single arginine residue, in
the attached pqr file. The command "molsurf test8.pqr 0.0" gives the
error message above.

I'm hoping that Paul Beroza (the author of molsurf) can figure this one
out. But anyone else on this list is welcome to give this a go as well!

> By the way, sander can calculate out the surface area of this
> structure by
> setting GB=1,PB=1,MS=0 in MM/PBSA input file. Results(SA value) from
> sander and molsurf seem identical, I'm wondering what's difference
> this two methods. Thanks!

I beleive that sander is using a numerical procedure based on
icosahedral partitioning of the vdW surface. This is fine for the SAS
(solvent accessible surface), but does not give the molecule surface
(SES), and is not analytical.

Holger (or someone else) can correct me if I am wrong here....


The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Fri Jul 15 2005 - 00:53:00 PDT
Custom Search