[gmx-users] Water organic solvents mixtures: Which force field to use and best practice to derive parameters

Thu Jun 4 21:08:43 CEST 2015

I just finished to write a tutorial on how to parameterize an organic molecule in OPLS/AA using MKTop, TopolGen, ACPYPE, RED Server, antechamber and Schrodinger Maestro. I also added a part to valid the charges and the final topology. Any comments, corrections and improvements are more than welcome.

http://www.esi.umontreal.ca/~pelletjo/gromacs/derive01.html

I reference Justin’s tutorials a lot. Thanks again to you for taking so much time and share them with the community. You make GROMACS so much more accessible for new users.

Max

On May 29, 2015, at 2:53 PM, Ebert Maximilian <m.ebert at umontreal.ca<mailto:m.ebert at umontreal.ca>> wrote:

I continued to define good partial charges using the tools suggested. I found that the combination of Maestro + ffconv or mktop, ACPYPE and antechamber + mktop have all advantages and disadvantages depending on the organic molecule.

for instance for acetone with maestro and mktop i could achieve the same topology as virtual chemistry. I next tried to generate a topology for iso-propanol. Here ACPYPE charges with the mktop atom types gave the closes density to the experimental one. I also tried to calculate the heat capacity with Cp and Cv to compare it with the literature. My box is 2.3x2.3x2.3nm large and has about 150 molecules of the organic solvent. After 1ns of NPT i get values which are 5x higher than the literature value (gmx energy -f npt.edr -fluct_props -nmol 100). I also tried a box 5x5x5nm with over 1000 molecules and got the same result. Any idea why the simulation using OPLS AA FF is so far of when it comes to the heat capacity?

thanks
max

On May 28, 2015, at 9:19 AM, Ebert Maximilian <m.ebert at umontreal.ca<mailto:m.ebert at umontreal.ca>> wrote:

Thanks Justin and Kalev this brings me already much further. I tried ffld_server and it works just fine. However, it is like a black box. I can’t really find the documentation on how ffld_server gets the charges. Do you know where to find the documentation?

Thanks

On May 28, 2015, at 2:22 AM, Kalev Takkis <kalev.takkis at gmail.com<mailto:kalev.takkis at gmail.com>> wrote:

If you're after OPLS topologies for GROMACS then one way to derive them is
via Schrödinger's Maestro (free academics version is sufficient) and
Andrey Frolov's
ffconv script (http://frolov-pchem.wikispaces.com/ffconv.py). You can
create a force field represesentation of a molecule with the former
(described here http://www.schrodinger.com/kb/809) and then convert it to
GROMACS format with the latter.

All the best,
Kalev

On 28 May 2015 at 03:37, Mohd Farid Ismail <mohd.farid.ismail at yandex.com<mailto:mohd.farid.ismail at yandex.com>>
wrote:

You can try R.E.D. Server.  It has more charge models (I don't know
whether that will help).

Also, IMO, one should target the density and the static dielectric
constant when it comes to VDW and partial charges.  I saw a recent paper
that might be of interest to you
http://pubs.acs.org/doi/abs/10.1021/jp3002383

--
Mohd Farid Ismail

28.05.2015, 05:13, "Ebert Maximilian" <m.ebert at umontreal.ca>:

I just finished a 1 ns NPT calculation of a 2.3x2.3x2.3 nm box filled with
acetone (130 molecules). The expected density at 300K is 784.1 kg/m^3. For
the virtual chemistry parameters i calculated 798.6 (close to the 800.1±0.2
value on their website) and for the parameter derived as explain in
previous mail I got 817.0 which seems too high. Does anybody has an advice
how I could improve the derivation of my parameters?

Thank you very much,

max

On May 27, 2015, at 3:25 PM, Ebert Maximilian <m.ebert at umontreal.ca>
wrote:

I read more about organic solvents in MD and came to the conclusion that
OPLS is indeed the best way to go. Since I couldn’t really find an
accessible tutorial how to derive topology files for GROMACS and the FF
OPLS/AA I will document my progress here. Maybe this is of help for
somebody in the future. In addition, I would like to ask the community to
help me in case you see problems with my approach. Once I have a good
protocol I will write a tutorial and make it available online.

To validate my approach I am trying to create a parameter set for acetone
which I found on  http://virtualchemistry.org. To generate the OPLS
topology I used a tool suggested by many people called mktop in version
2.2.1. I downloaded the ideal geometry of acetone from Ligand Expo and
generated a GROMACS topology file using the following command:

mktop_2.2.1.pl -i ACN_ideal.pdb -o acn_topology.top -ff opls -conect yes

In order to get the charges for this organic molecule I downloaded the
most recent amber tools and compiled it. I used the AM1-BCC charge model to
generate charges for acetone using the following instructions in
antechamber:

antechamber -i ACN_ideal.pdb -fi pdb -o acn.mol2 -fo mol2 -c bcc -s 2

I opened the resulting mol2 file in Chimera to map the atoms to the atoms
in my .top file. The charges calculated by antechamber look reasonable and
are comparable to the validated OPLS topology from virtual chemistry:

virtual chemistry charges

[ atoms ]
;   nr       type  resnr residue  atom   cgnr     charge       mass
typeB    chargeB      massB
      1  opls_280         1       LIG         C         1      0.47
12.011
      2  opls_135         1       LIG         C         2     -0.18
12.011
      3  opls_135         1       LIG         C         3     -0.18
12.011
      4  opls_281         1       LIG         O         4     -0.47
15.9994
      5  opls_282         1       LIG         H         5      0.06
 1.008
      6  opls_282         1       LIG         H         6      0.06
 1.008
      7  opls_282         1       LIG         H         7      0.06
 1.008
      8  opls_282         1       LIG         H         8      0.06
 1.008
      9  opls_282         1       LIG         H         9      0.06
 1.008
     10  opls_282         1       LIG         H        10      0.06
 1.008

antechamber AM1-BCC derived

[ atoms ]
;   nr       type  resnr residue  atom   cgnr     charge       mass
typeB    chargeB      massB
     1  opls_280   1   ACN      C1    1    0.56     12.011
     2  opls_281   1   ACN      O1    1   -0.52     15.9994
     3  opls_135   1   ACN      C2    2   -0.20     12.011
     4  opls_135   1   ACN      C3    3   -0.20     12.011
     5  opls_282   1   ACN      H1    2    0.06     1.008
     6  opls_282   1   ACN      H2    2    0.06     1.008
     7  opls_282   1   ACN      H3    2    0.06     1.008
     8  opls_282   1   ACN      H4    3    0.06     1.008
     9  opls_282   1   ACN      H5    3    0.06     1.008
    10  opls_282   1   ACN      H6    3    0.06     1.008

The atom types were guessed correctly by mktop and also the charge groups
make sense I think. So far so good.

I realize some differences between the two topologies. First the mktop
topology also includes FF constants for the different bonds and angles:

[ bonds ]
1 2 1   0.121  476976.0
1 3 1   0.151  265265.6
1 4 1   0.151  265265.6
3 5 1   0.109  284512.0
3 6 1   0.109  284512.0
3 7 1   0.109  284512.0
4 8 1   0.109  284512.0
4 9 1   0.109  284512.0
4 10 1   0.109  284512.0

[ angles ]
1 3 5 1  109.460  292.880
1 3 6 1  109.473  292.880
1 3 7 1  109.484  292.880
1 4 8 1  109.466  292.880
1 4 9 1  109.435  292.880
1 4 10 1  109.477  292.880
2 1 3 1  119.985  669.440
2 1 4 1  119.985  669.440
3 1 4 1  120.029  585.760
5 3 6 1  109.445  276.144
5 3 7 1  109.464  276.144
6 3 7 1  109.502  276.144
8 4 9 1  109.483  276.144
8 4 10 1  109.504  276.144
9 4 10 1  109.462  276.144

compared to the virtual chemistry file:

[ bonds ]
;  ai    aj funct            c0            c1            c2            c3
 1     2     1
 1     3     1
 1     4     1
 2     5     1
 2     6     1
 2     7     1
 3     8     1
 3     9     1
 3    10     1

[ angles ]
;  ai    aj    ak funct            c0            c1            c2
        c3
 2     1     3     1
 2     1     4     1
 3     1     4     1
 1     2     5     1
 1     2     6     1
 1     2     7     1
 5     2     6     1
 5     2     7     1
 6     2     7     1
 1     3     8     1
 1     3     9     1
 1     3    10     1
 8     3     9     1
 8     3    10     1
 9     3    10     1

Should I trust the mktop parameters or delete them? To look if my
parameters are correct I did a short MD with a box containing only acetone
based on the two topologies. The MD is still running but I wanted to
compare the density and see how it matches with reality.

What do you think about this approach? What would have been a better way?
How can I make sure that the charges are correct?

Thanks for your input.

Max

On May 27, 2015, at 11:54 AM, Ebert Maximilian <m.ebert at umontreal.ca
<mailto:m.ebert at umontreal.ca>> wrote:

Hi there,

I am about to setup a water:organic solvent mixture with a protein. I
found many organic molecules on http://virtualchemistry.org with
definitions for the OPLS FF. However, some are missing so I would need to
derive the parameters myself. Before going into more details I was
wondering if OPLS is to be preferred if organic solvent is present or can
AMBER also be used? It seems that using ACPYPE with AMBER is much more
accessible than using any other method to derive the parameters for organic
molecules.

Thanks for your advice.
--
Gromacs Users mailing list

* Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
posting!