ForceBalance and Amber

[Isaiah Sumner]

Hi,
I'm sure there is a simple answer to this question. I am trying to use force balance to force match a force field using AmberMD as an engine. It seems that the code is able to handle this, at least according to the manual. I started with the studies/001_water_tutorial, and changed the input file, very_simple.in and changed the following line:

simtype abinitio_gmx

to

simtype AbInitio_AMBER

I assumed this would give me an error eventually, but when I ran "ForceBalance.py very_simple.in" I got the following error, which was unexpected:

/usr/local/lib/python2.7/site-packages/forcebalance/contact.py:58: FutureWarning: comparison to `None` will result in an elementwise object comparison in the future.
if box == None:
Traceback (most recent call last):
File "/usr/local/bin/ForceBalance.py", line 39, in Run_ForceBalance
objective = Objective(options, tgt_opts, forcefield)
File "/usr/local/lib/python2.7/site-packages/forcebalance/objective.py", line 134, in __init__
else: Tgt = Implemented_Targets[opts['type']](options,opts,forcefield)
File "/usr/local/lib/python2.7/site-packages/forcebalance/amberio.py", line 173, in __init__
super(AbInitio_AMBER,self).__init__(options,tgt_opts,forcefield)
File "/usr/local/lib/python2.7/site-packages/forcebalance/abinitio.py", line 155, in __init__
self.engine = self.engine_(target=self, mol=self.mol, **engine_args)
AttributeError: 'AbInitio_AMBER' object has no attribute 'engine_'

The sticking point seems to be the last part. I'm not sure where to go from here, so any insight would be appreciated.

Thanks for your help,

Isaiah

[Lee-Ping Wang]

Hi Isaiah,

Thanks for your message. I'm glad you're interested in using ForceBalance. ForceBalance does support AMBER for parameterizing force fields, but it can't convert a GROMACS force field format to AMBER. If you change AbInitio_GMX to AbInitio_AMBER, you also need to change some files in forcefield and targets.

I've just updated the master branch with an example (studies/014_chromophore) that does energy + force matching along with RESP in AMBER. The molecule is parameterized starting from a .mol2 and .frcmod file that you can create using Antechamber. The files in "targets" are generated from quantum chemistry calculations.

Let me know if this works for you.

Thanks,

- Lee-Ping

[Isaiah Sumner]

Hi Lee-Ping,

Thanks for your reply! I'll test it out and let you know how it goes.

Isaiah

[Isaiah Sumner]

Hi Lee-Ping,
Let me preface this by saying that I am not a python programmer. So, I'm sure this is another easy fix. When I downloaded the latest branch with your example (/usr/local/bin/ForceBalance optimize.in), I get the following error:

File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/amberio.py", line 995, in energy_force
return np.hstack((Result["Energy"].reshape(-1,1), Result["Force"]))
File "/usr/local/lib/python2.7/site-packages/numpy/core/shape_base.py", line 277, in hstack
return _nx.concatenate(arrs, 1)
ValueError: all the input arrays must have same number of dimensions

I can "fix" this (again, not a python programmer so it's possible this was wrong) by making the following change in amberio.py:

return np.hstack((Result["Energy"].reshape(-1,1), Result["Force"])) ====> return np.column_stack((Result["Energy"].reshape(-1,1), Result["Force"]))

I rerun it, and get a new error. This error is an amber error, I believe. Amber will not overwrite forcedump.dat. I fixed that error by putting in the following command:

os.system("rm -f forcedump.dat")

just before Amber is called (self.callamber("sander...

This results in this error:

Traceback (most recent call last):
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/EGG-INFO/scripts/ForceBalance.py", line 43, in Run_ForceBalance
optimizer.Run()
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/optimizer.py", line 305, in Run
xk = self.OptTab[self.jobtype]()
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/optimizer.py", line 887, in NewtonRaphson
return self.MainOptimizer(b_BFGS=0)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/optimizer.py", line 471, in MainOptimizer
data = self.Objective.Full(xk,Ord,verbose=True)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/objective.py", line 271, in Full
Objective = self.Target_Terms(mvals, Order, verbose, customdir)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/objective.py", line 213, in Target_Terms
Ans = Funcs[Order](mvals, customdir=customdir)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/target.py", line 310, in get_H
Ans = self.meta_get(mvals,1,1,customdir=customdir)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/target.py", line 565, in meta_get
Answer = self.get(mvals, AGrad, AHess)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/abinitio.py", line 1281, in get
Answer_EF = self.get_energy_force(mvals, AGrad, AHess)
File "/usr/local/lib/python2.7/site-packages/forcebalance-1.3.2-py2.7-linux-x86_64.egg/forcebalance/abinitio.py", line 771, in get_energy_force
dM_all[:,p,:], ddM_all[:,p,:] = f12d3p(fdwrap(callM, mvals, p), h = self.h, f0 = M_all)
ValueError: could not broadcast input array from shape (0,211) into shape (100,211)

I don't know where to go from here. I am using Python 2.7.6 and the current version of the code from github.

Please let me know if you want or need more information.

Thanks for your help,

Isaiah

[Lee-Ping Wang]

Ahh, I think I know what's going on. Thanks for the error report.

With the current AMBER version, I don't think it correctly loops over the whole trajectory when writing the forces; instead it just writes the first force and quits. Additionally, it prints out the force components (i.e. bond, angle etc.) which ForceBalance does not need. I worked around this by patching the sander source and recompiling it, but I have yet to document it.

The patch file is here: https://dl.dropboxusercontent.com/u/538 ... -patch.txt

To apply the patch, make sure you're using AmberTools 15 (the newest version), go into the $AMBERHOME folder (containing AmberTools) and type "patch -p1 < force-patch.txt".

You should see this output:

Code: Select all

leeping@leeping:~/opt/amber$ patch -p1 < force-patch.txt
patching file AmberTools/src/sander/debug.F90
patching file AmberTools/src/sander/trajene.F90

After recompiling AmberTools (specifically sander), the error should go away.

Just to be sure: Can you look at the dimensions of the array to ensure they are really incompatible? There's supposed to be a 1-D array for the energy containing "ns" elements (i.e. number of snapshots), and a 2D array for the energy containing (ns, 3N) elements.

[Isaiah Sumner]

It seems to be working now. I'll poke around and get you the other information you requested. If I manage to break it again, I'll let you know.

Thanks,

Isaiah

[Lee-Ping Wang]

Glad to hear it. Yes, please stay in touch.

- Lee-Ping