[AMBER-Developers] Discussion towards removing tip3p parameters from protein, RNA & DNA ff

From: Kellon Belfon <kellonbelfon.gmail.com>
Date: Tue, 9 Mar 2021 11:32:59 -0500

Hi Everyone,

Towards the end of the Amber Dev meeting, there was a small discussion led
by Alexey on removing tip3p parameters from the protein, RNA, and DNA force
fields due to the problem of them overwriting opc parameters when they are
sourced after opc leaprc (leaprc.water.opc.)

Based on what I observed, each leaprc file that corresponds to these force
fields uses some version of the parm10.dat file (e.g parm19.dat for 19SB
and parm15ipq_10.3.dat for IPQ, parm99.dat for some protein ff, etc). In
those files, the HW and OW parameters for tip3p are present.

Therefore I started my test by removing all mentions of HW and OW
parameters from these parmxx.dat files. By doing this approach users will
have to explicitly source the tip3p leaprc (leaprc.water.tip3p) in order to
get the HW and OW parameters for tip3p. Of course, as expected this will
break a few test cases that relied on getting the tip3p parameters from
only sourcing the protein, RNA or DNA force fields.

To summarize, 32 tests failed in AmberTools. Those tests correspond to
failures in the 1) leap test suite, 2) pymsmt test suite, and 3) FEW test

*Failures can be categorized in the following types:*

   1. Differences in NTYPES and NPHB in the topology generated from having
   or not having tip3p parameters in the protein/RNA/DNA forcefields.
   2. Missing HW/OW parameters resulting in failure to generate topology
   files causing failed diffs of leap.log and parm7 files.

   - NTYPES: total number of distinct atom types
   - NPHB : number of distinct 10-12 hydrogen bond pair types

In the case of failure type 2, adding the line source water.leaprc.tip3p in
the test case leap.in file, fixes the missing parameters error but results
in failure type 1 error.

The differences in NTYPES is due to how the HW atom type is categorized.
When I source water.leaprc.tip3p and remove the tip3p parameters from the
parmxx.dat files, the HW atom type is placed in the same category as the HO
atom type. When the tip3p parameters are read from the parmxx.dat files
(current method) then the HW atom type is placed in its own category. This
results in the topology files generated from the old test cases having one
more NTYPES and NPHB = 1 than the new topology files (removing tip3p
parameters from parmxx.dat files). In certain tests the NTYPES are the same
but NPHB differs once again (1 vs 0). I am not sure exactly where leap
categorizes the atom types. I can spend time looking at the code but I
figure someone might know about this and can provide some insights.

*Solutions/Proposal/Questions: *

   - If anyone else does not object (big possibility I am missing some
   knowledge on how things were done in amber in the past) then I propose to
   regenerate the topology and leap log files for the test failures with
   NTYPES and NPHB differences.

   - I also proposed to change the leap.in for the tests that fail to
   source leaprc.water.tip3p.

   - For the old forcefields (oldff directory in cmd directory) that loads
   water and ion parameters along with the parmxx.dat file, should we remove
   the loading of the water and ion parameters and source tip3p in place to
   keep it from breaking users inputs that use older forcefield? This should
   not be a concern for the new force fields since sourcing a water model is
   part of the protocol.

   - For the pymsmt test failures, I cannot find a leap.in file to change.
   Does the mcpb.py script generate its own leap.in files? Is there a
   specific reason for using oldff/leaprc.ff14SB vs leaprc.protein.ff14SB for
   some of the pymsmt tests?

*Unique case: *
One of the test case $AMBERHOME/AmberTools/test/leap/Run.cmds tend to give
different water coordinates and number of water molecules when source
leaprc.water.tip3p vs keeping the tip3p parameters in the parm10.dat file

I can spend time trying to figure out what is going on above but I figured
I share it here and maybe someone might know "off the top of their head".

I look forward to a discussion on this and coming to a conclusion on what
the best move forward on handling this situation.

If anyone wants a breakdown of the test failures and the parmxx.dat files
that each leaprc in amber loads, I have this information that I can share.

Thank you.

Kellon A. A. Belfon, Graduate Student
Carlos Simmerling Laboratory
The Laufer Center for Physical and Quantitative Biology
The Department of Chemistry, Stony Brook University
Stony Brook, New York 11794
Phone: (347) 546-4237 <(347)+546+4237>  Email:  kellon.belfon.stonybrook.
AMBER-Developers mailing list
Received on Tue Mar 09 2021 - 09:00:02 PST
Custom Search