[AMBER] PBSA in AT14: invalid test case, estatic potential values are written in wrong units to DX files.

From: Jan-Philip Gehrcke <jgehrcke.googlemail.com>
Date: Mon, 08 Sep 2014 15:57:00 +0200

Hello!

There is this daunting problem in PBSA affecting me since three releases
(AT12, 13, 14). Let's fix this, once and for all, please :).

Short story, what needs to be done:

1) In AmberTools/src/pbsa/pb_force.F90, line 1438, change

     call gen_dx_file(xm,ym,zm,h,gox,goy,goz,phi(1:xmymzm),&

to

     call gen_dx_file(xm,ym,zm,h,gox,goy,goz,frcfac*phi(1:xmymzm),&

2) Adjust test case AmberTools/test/pbsa_phiout.


Long story:

More than two years ago, I reported that electrostatic potential values
are written to a DX file by PBSA in wrong units ("internal units"
instead of kcal/(mol*e): http://archive.ambermd.org/201207/0466.html

This was then fixed in AT12
(http://ambermd.org/bugfixes/AmberTools/12.0/bugfix.17). The problem was
re-introduced in AT13, and I reported again:
http://archive.ambermd.org/201311/0380.html

A fix for AT13 has unfortunately never been released (according to
http://ambermd.org/bugfixesat.html).

My hope now was on AT 14. A fix for AT14 is documented with "update.5:
Fix a number of small bugs in PBSA" ... and "Potential printing was in
the internal unit."

Now that I compiled it and tried it out, I realized that PBSA is still
behaving erroneously. The facts:

- A test case was added, but someone did not properly check the numbers,
and included a test case with the invalid values being the "golden
standard".
- The changelog of AT14 update.5 is out-of-sync with the actual code
changes.

How to verify that there is a problem:

1) Go to directory amber14/AmberTools/test/pbsa_phiout.

2) View file pbsa_phi.phi.save. The absolute value of most of the
numbers is below 1. Particularly large absolute values are on the order
of 10.

3) View pbsa_phi.dx.save. The absolute value of most of the numbers is
on the order of 100. Particularly large absolute values are on the order
of 1000.

These are just two different file formats that should show the same
numeric values. Actually, I have found this by investigating the system
of which I know that electrostatic potential values of about 10
kcal/(mol*e) are realistic, but the DX file just contains too large
numbers. And indeed, the pb_force.F90 part of
http://ambermd.org/bugfixes/AmberTools/12.0/bugfix.17 is just not
applied in AT14. so it makes sense that we are back to old (erroneous)
behavior.


Hoping for a PBSA that is working out-of-the-box for my purposes.. :)

Cheers!

Jan-Philip











_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Sep 08 2014 - 07:00:07 PDT
Custom Search