[gmx-users] AdResS feature in gromacs

[Jan Henning Peters]

Hello Eduardo,

I forgot that water of course does not require 1-4 interactions, so 
that's while the tf generation worked (VOTCA would not have generated a 
tablep.xvg file if it had been required but already caused problems 
there) and the simulation with protein didn't. Again, copying the 
table.xvg (or the file you used) should solve this problem.

As long as the protein does not enter the transition region, you should 
be able to use the tf calculated for a pure water box, but you should 
check that the results are reasonable (no change in water density in the 
hybrid region). However, you should take care when adapting the tf table 
- if you are using a bigger explicit region (adress-ex-width in the .mdp 
file), just extending the tf table at the end (I believe that is what 
VOTCA would do) will not do the trick, but you will have to shift the 
table. For example, if you calculated the tf for adress-ex-width = 2 and 
now use adress-ex-width = 5, you should add 3 to the first column (r) of 
the table and add lines for r<3 with constant second (the potential 
U(r<3) = U(r=3)) and third (the force F(r<3) = 0.0) columns. I would not 
recommend changing the width of the hybrid region (adress-hy-width in 
the .mdp file) - it might be possible to adapt the tf, but it's better 
if you don't.

The AdResS region in GROMACS does not move but always is centred around 
the reference point (adress-reference-coords in the .mdp file), so you 
will have to make sure that this reference point is - and remains - 
close to the centre of the protein. Currently, the "usual" (not many 
people have done something like this) approach for this would be to 
apply position restraints to parts of the solute (ideally only one atom) 
- there are approaches for mobile explicit regions, but they are not 
implemented in GROMACS. Also, you should make sure that the explicit 
region is big enough that no part of the protein leaves the explicit 
region (after all, the protein is not static). If parts of the protein 
leave the explicit region, the simulation will likely continue for some 
time before running into problems, but the results are meaningless, even 
if no errors occur.

By the way, you might want to look at this publication, where something 
similar was done for micelles: https://doi.org/10.1002/adts.201800025

Regards,
Jan

On 03/07/2018 18:13, Eduardo Diniz wrote:
> Hello Jan
>
> Thank you for your reply
>
> Exactly, VOTCA generated the thermoforce so it was able to perform an Adress simulation.
> I've looked more carefully at the files generated by VOTCA and could not find a file that
> might be this tablep.xvg. But I think that this problem is related to the
> fact that my sistem have a protein. The system used for VOTCA to generate the thermoforce
> contained only water molecules. I thought at first the the thermoforce generated for a system of
> pure water could be transferred to other systems containing a solute. I don't know if this is correct.
> I tried your suggestion and copied a 6-12 potential from gromacs/share/top/table6-12.xvg to a
> tablep.xvg file and gmx mdrun returned:
> AdResS full box therm force table in file tabletf.xvg extends to 7.100000:
>      should extend to at least for spherical adress13.569794 (=distance from center to furthermost point in box
>
> which is related to the extension of my thermodynamic force table. This same table extension worked well for generating the
> thermoforce with VOTCA. I don't know if the spheric atomistic region is attached
> to some atom of the protein so it moves with the protein or if it is a static region, in which case the protein
> would cross the transition region, what could explain the need for this tablep.xvg I think.
>
> Please, if you could help I would really appreciate.
> ____________________________________________________________________________________
> Hello Eduardo,
>
> tablep.xvg contains the tabulated potential for 1-4 interactions, which
> are non-bonded interactions inside a molecule that usually only differ
> by the Lennard-Jones parameters sigma/epsilon (or C6/C12) but not the
> general shape of the potential. Since these should only occur in the
> fine-grained part, they should just be standard 6-12 Lennard-Jones
> potentials and you could just copy table.xvg (the 6-12 potential used
> for fine-grained non-bonded interactions) to tablep.xvg.
>
> However, since you used VOTCA to generate the tabletf.xvg, it did
> successfully run AdResS simulations, so my guess is that you already
> have proper tablep.xvg files that were used in the tf generation (I
> would also assume that they are indeed identical to the table.xvg files
> used). Hence you could also copy a tablep.xvg file from one of the
> step_* subdirectories of your tf calculation run.
>
> Regards,
>
> Jan
>
>
> On 03/07/2018 15:24, Eduardo Diniz wrote:
>> Hello everyone,
>>
>>
>> I'm trying to run an simulation in gromacs with AdResS, where water is treated as fine grained in the region close to the protein and as coarse-grained in the farthest layers. I've already generated the tabulated potential for coarse-grained water with the VOTCA package and also the termodynamic force table with the same package. I generated the tpr file with no error and when I try to run mdrun the command complains about a missing file:
>>
>>
>> System I/O error:
>> Library file 'tablep.xvg' not found in current dir nor in the default
>> directories.
>> The following paths were searched:
>>
>> I have no idea what is this tablep.xvg file. There is no problem in finding the thermodynamic force (tabletf.xvg) nor the coarse-grained tabulated potential (table.xvg). I tested it by removing this files, the command shows a similar message but instead complaining about the table.xvg/tabletf.xvg missing file. I appreciate if anyone could help me.
>
>
> ________________________________
>