Re: [gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[Mark Abraham]

Hi,

Best practice is to read and learn others practice from publications that
are similar to what you want to do, rather than making ad-hoc changes. In
this case, the GROMACS defaults are pretty close to the de facto standard,
and supported by analysis work done by other members of the community.

Mark

On Thu, Jun 30, 2016 at 4:16 PM NISHA Prakash 
wrote:

> Dear Justin,
>
> Thanks a lot for pointing out the issues. I now understand why there were
> such high oscillations.
>
> Could you please also tell me if there are any ideal values for pme_order
> and fourier spacing with respect to the cut offs' value of 1.4.
>
> Does the following Note imply I can raise the fourier grid spacing to 0.25?
>
> NOTE 2 [file sim-new.mdp]:
>   The optimal PME mesh load for parallel simulations is below 0.5
>   and for highly parallel simulations between 0.25 and 0.33,
>   for higher performance, increase the cut-off and the PME grid spacing
>
> Thank you again,
>
> Nisha
>
>
> On Thu, Jun 30, 2016 at 6:55 PM, Justin Lemkul  wrote:
>
> >
> >
> > On 6/30/16 9:16 AM, NISHA Prakash wrote:
> >
> >> Dear Justin,
> >>
> >> Thank you for your reply.
> >> It is a protein carbohydrate system.  Including the solvent, the number
> of
> >> atoms is 43499.
> >> I have minimized the system for 200 ps followed by NPT and NVT
> simulations
> >> for 200 ps respectively
> >>
> >>
> > Given that your temperature output started from 0 K, then you did not
> > continue from the equilibration properly by supplying the checkpoint file
> > to grompp -t. This is important to get right, otherwise you're basically
> > starting over from some random point (an equilibrated structure without
> any
> > velocities likely isn't a physically realistic state).
> >
> > Below is the .mdp file.
> >>
> >>
> >> ; VARIOUS PREPROCESSING OPTIONS
> >> title= REMD Simulation
> >> define   = -DPOSRES
> >>
> >>
> >> ; RUN CONTROL PARAMETERS
> >> integrator   = md-vv  ; velocity verlet algorithm -
> >> tinit= 0 ;
> >> dt   = 0.002; timestep in ps
> >> nsteps  = 500;
> >> simulation-part  = 1 ; Part index is updated automatically on
> >> checkpointing
> >> comm-mode= Linear ; mode for center of mass motion
> removal
> >> nstcomm  = 100 ; number of steps for center of mass
> motion
> >> removal
> >> comm-grps= Protein_Carb  Water_and_Ions ; group(s)
> for
> >> center of mass motion removal
> >>
> >>
> > In a solvated system, you should not be separating these groups.  This
> > could explain the sudden jump in temperature - you could have things
> > clashing badly over the course of the simulation.
> >
> >
> >
> >> ; ENERGY MINIMIZATION OPTIONS
> >> emtol= 10 ; Force tolerance
> >> emstep   = 0.01 ; initial step-size
> >> niter= 20 ; Max number of iterations in relax-shells
> >> fcstep   = 0 ; Step size (ps^2) for minimization of
> >> flexible constraints
> >> nstcgsteep   = 1000 ; Frequency of steepest descents steps
> >> when
> >> doing CG
> >> nbfgscorr= 10
> >>
> >>
> >> ; OUTPUT CONTROL OPTIONS
> >> nstxout  = 5 ; Writing full precision coordinates
> >> every
> >> ns
> >> nstvout  = 5 ; Writing velocities every nanosecond
> >> nstfout  = 0 ; Not writing forces
> >> nstlog   = 5000  ; Writing to the log file every step
> 10ps
> >> nstcalcenergy= 100
> >> nstenergy= 5000  ; Writing out energy information every
> >> step 10ps
> >> nstxtcout= 2500  ; Writing coordinates every 5 ps
> >> xtc-precision= 1000
> >> xtc-grps = Protein_Carb  Water_and_Ions ; subset of
> >> atoms for the .xtc file.
> >> energygrps   = Protein_Carb  Water_and_Ions ; Selection
> of
> >> energy groups
> >>
> >>
> >> ; NEIGHBORSEARCHING PARAMETERS
> >> nstlist  = 10 ; nblist update frequency-
> >> ns-type  = Grid ; ns algorithm (simple or grid)
> >> pbc  = xyz ; Periodic boundary conditions: xyz,
> >> no,
> >> xy
> >> periodic-molecules   = no
> >> rlist= 1.4 ;  nblist cut-off
> >> rlistlong= -1 ; long-range cut-off for switched
> >> potentials
> >>
> >>
> >> ; OPTIONS FOR ELECTROSTATICS
> >> coulombtype  = PME ; Method for doing electrostatics
> >> rcoulomb = 1.4 ;
> >> epsilon-r= 1 ; Relative dielectric constant for the
> >> medium
> >> pme_order= 10;
> >>
> >>
> >> ; OPTIONS FOR VDW
> >> vdw-type = Cut-off  ; Method for doing Van der Waals
> >> rvdw-switch  = 0 ; cut-off lengths
> >> rvdw = 1.4 ;

Re: [gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[NISHA Prakash]

Dear Justin,

Thanks a lot for pointing out the issues. I now understand why there were
such high oscillations.

Could you please also tell me if there are any ideal values for pme_order
and fourier spacing with respect to the cut offs' value of 1.4.

Does the following Note imply I can raise the fourier grid spacing to 0.25?

NOTE 2 [file sim-new.mdp]:
  The optimal PME mesh load for parallel simulations is below 0.5
  and for highly parallel simulations between 0.25 and 0.33,
  for higher performance, increase the cut-off and the PME grid spacing

Thank you again,

Nisha


On Thu, Jun 30, 2016 at 6:55 PM, Justin Lemkul  wrote:

>
>
> On 6/30/16 9:16 AM, NISHA Prakash wrote:
>
>> Dear Justin,
>>
>> Thank you for your reply.
>> It is a protein carbohydrate system.  Including the solvent, the number of
>> atoms is 43499.
>> I have minimized the system for 200 ps followed by NPT and NVT simulations
>> for 200 ps respectively
>>
>>
> Given that your temperature output started from 0 K, then you did not
> continue from the equilibration properly by supplying the checkpoint file
> to grompp -t. This is important to get right, otherwise you're basically
> starting over from some random point (an equilibrated structure without any
> velocities likely isn't a physically realistic state).
>
> Below is the .mdp file.
>>
>>
>> ; VARIOUS PREPROCESSING OPTIONS
>> title= REMD Simulation
>> define   = -DPOSRES
>>
>>
>> ; RUN CONTROL PARAMETERS
>> integrator   = md-vv  ; velocity verlet algorithm -
>> tinit= 0 ;
>> dt   = 0.002; timestep in ps
>> nsteps  = 500;
>> simulation-part  = 1 ; Part index is updated automatically on
>> checkpointing
>> comm-mode= Linear ; mode for center of mass motion removal
>> nstcomm  = 100 ; number of steps for center of mass motion
>> removal
>> comm-grps= Protein_Carb  Water_and_Ions ; group(s) for
>> center of mass motion removal
>>
>>
> In a solvated system, you should not be separating these groups.  This
> could explain the sudden jump in temperature - you could have things
> clashing badly over the course of the simulation.
>
>
>
>> ; ENERGY MINIMIZATION OPTIONS
>> emtol= 10 ; Force tolerance
>> emstep   = 0.01 ; initial step-size
>> niter= 20 ; Max number of iterations in relax-shells
>> fcstep   = 0 ; Step size (ps^2) for minimization of
>> flexible constraints
>> nstcgsteep   = 1000 ; Frequency of steepest descents steps
>> when
>> doing CG
>> nbfgscorr= 10
>>
>>
>> ; OUTPUT CONTROL OPTIONS
>> nstxout  = 5 ; Writing full precision coordinates
>> every
>> ns
>> nstvout  = 5 ; Writing velocities every nanosecond
>> nstfout  = 0 ; Not writing forces
>> nstlog   = 5000  ; Writing to the log file every step 10ps
>> nstcalcenergy= 100
>> nstenergy= 5000  ; Writing out energy information every
>> step 10ps
>> nstxtcout= 2500  ; Writing coordinates every 5 ps
>> xtc-precision= 1000
>> xtc-grps = Protein_Carb  Water_and_Ions ; subset of
>> atoms for the .xtc file.
>> energygrps   = Protein_Carb  Water_and_Ions ; Selection of
>> energy groups
>>
>>
>> ; NEIGHBORSEARCHING PARAMETERS
>> nstlist  = 10 ; nblist update frequency-
>> ns-type  = Grid ; ns algorithm (simple or grid)
>> pbc  = xyz ; Periodic boundary conditions: xyz,
>> no,
>> xy
>> periodic-molecules   = no
>> rlist= 1.4 ;  nblist cut-off
>> rlistlong= -1 ; long-range cut-off for switched
>> potentials
>>
>>
>> ; OPTIONS FOR ELECTROSTATICS
>> coulombtype  = PME ; Method for doing electrostatics
>> rcoulomb = 1.4 ;
>> epsilon-r= 1 ; Relative dielectric constant for the
>> medium
>> pme_order= 10;
>>
>>
>> ; OPTIONS FOR VDW
>> vdw-type = Cut-off  ; Method for doing Van der Waals
>> rvdw-switch  = 0 ; cut-off lengths
>> rvdw = 1.4 ;
>> DispCorr = EnerPres; Apply long range dispersion
>> corrections for Energy and Pressure
>> table-extension  = 1; Extension of the potential lookup tables
>> beyond the cut-off
>> fourierspacing   = 0.08;  Spacing for the PME/PPPM FFT grid
>>
>>
> This small Fourier spacing, coupled with the very high PME order above, is
> going to unnecessarily slow your system down.  Is there some reason you
> have set these this way?
>
>
>> ; GENERALIZED BORN ELECTROSTATICS
>> gb-algorithm = Still; Algorithm for calculating Born radii
>> nstgbradii   = 1; Frequency of 

Re: [gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[Justin Lemkul]

On 6/30/16 9:16 AM, NISHA Prakash wrote:

Dear Justin,

Thank you for your reply.
It is a protein carbohydrate system.  Including the solvent, the number of
atoms is 43499.
I have minimized the system for 200 ps followed by NPT and NVT simulations
for 200 ps respectively



Given that your temperature output started from 0 K, then you did not continue 
from the equilibration properly by supplying the checkpoint file to grompp -t. 
This is important to get right, otherwise you're basically starting over from 
some random point (an equilibrated structure without any velocities likely isn't 
a physically realistic state).



Below is the .mdp file.


; VARIOUS PREPROCESSING OPTIONS
title= REMD Simulation
define   = -DPOSRES


; RUN CONTROL PARAMETERS
integrator   = md-vv  ; velocity verlet algorithm -
tinit= 0 ;
dt   = 0.002; timestep in ps
nsteps  = 500;
simulation-part  = 1 ; Part index is updated automatically on
checkpointing
comm-mode= Linear ; mode for center of mass motion removal
nstcomm  = 100 ; number of steps for center of mass motion
removal
comm-grps= Protein_Carb  Water_and_Ions ; group(s) for
center of mass motion removal



In a solvated system, you should not be separating these groups.  This could 
explain the sudden jump in temperature - you could have things clashing badly 
over the course of the simulation.




; ENERGY MINIMIZATION OPTIONS
emtol= 10 ; Force tolerance
emstep   = 0.01 ; initial step-size
niter= 20 ; Max number of iterations in relax-shells
fcstep   = 0 ; Step size (ps^2) for minimization of
flexible constraints
nstcgsteep   = 1000 ; Frequency of steepest descents steps when
doing CG
nbfgscorr= 10


; OUTPUT CONTROL OPTIONS
nstxout  = 5 ; Writing full precision coordinates every
ns
nstvout  = 5 ; Writing velocities every nanosecond
nstfout  = 0 ; Not writing forces
nstlog   = 5000  ; Writing to the log file every step 10ps
nstcalcenergy= 100
nstenergy= 5000  ; Writing out energy information every
step 10ps
nstxtcout= 2500  ; Writing coordinates every 5 ps
xtc-precision= 1000
xtc-grps = Protein_Carb  Water_and_Ions ; subset of
atoms for the .xtc file.
energygrps   = Protein_Carb  Water_and_Ions ; Selection of
energy groups


; NEIGHBORSEARCHING PARAMETERS
nstlist  = 10 ; nblist update frequency-
ns-type  = Grid ; ns algorithm (simple or grid)
pbc  = xyz ; Periodic boundary conditions: xyz, no,
xy
periodic-molecules   = no
rlist= 1.4 ;  nblist cut-off
rlistlong= -1 ; long-range cut-off for switched
potentials


; OPTIONS FOR ELECTROSTATICS
coulombtype  = PME ; Method for doing electrostatics
rcoulomb = 1.4 ;
epsilon-r= 1 ; Relative dielectric constant for the
medium
pme_order= 10;


; OPTIONS FOR VDW
vdw-type = Cut-off  ; Method for doing Van der Waals
rvdw-switch  = 0 ; cut-off lengths
rvdw = 1.4 ;
DispCorr = EnerPres; Apply long range dispersion
corrections for Energy and Pressure
table-extension  = 1; Extension of the potential lookup tables
beyond the cut-off
fourierspacing   = 0.08;  Spacing for the PME/PPPM FFT grid



This small Fourier spacing, coupled with the very high PME order above, is going 
to unnecessarily slow your system down.  Is there some reason you have set these 
this way?




; GENERALIZED BORN ELECTROSTATICS
gb-algorithm = Still; Algorithm for calculating Born radii
nstgbradii   = 1; Frequency of calculating the Born radii
inside rlist
rgbradii = 1; Cutoff for Born radii calculation
gb-epsilon-solvent   = 80; Dielectric coefficient of the implicit
solvent
gb-saltconc  = 0; Salt concentration in M for Generalized
Born models


; Scaling factors used in the OBC GB model. Default values are OBC(II)
gb-obc-alpha = 1
gb-obc-beta  = 0.8
gb-obc-gamma = 4.85
gb-dielectric-offset = 0.009
sa-algorithm = Ace-approximation
sa-surface-tension   = -1; Surface tension (kJ/mol/nm^2) for the SA
(nonpolar surface) part of GBSA - default -1



Implicit solvent should not be used if you have explicit solvent, though it 
looks like these options are probably off since the default for the 
implicit-solvent keyword is "no," but be aware that these are extraneous.





; Temperature coupling
tcoupl = nose-hoover
nsttcouple  

Re: [gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[NISHA Prakash]

Dear Justin,

Thank you for your reply.
It is a protein carbohydrate system.  Including the solvent, the number of
atoms is 43499.
I have minimized the system for 200 ps followed by NPT and NVT simulations
for 200 ps respectively

Below is the .mdp file.


; VARIOUS PREPROCESSING OPTIONS
title= REMD Simulation
define   = -DPOSRES


; RUN CONTROL PARAMETERS
integrator   = md-vv  ; velocity verlet algorithm -
tinit= 0 ;
dt   = 0.002; timestep in ps
nsteps  = 500;
simulation-part  = 1 ; Part index is updated automatically on
checkpointing
comm-mode= Linear ; mode for center of mass motion removal
nstcomm  = 100 ; number of steps for center of mass motion
removal
comm-grps= Protein_Carb  Water_and_Ions ; group(s) for
center of mass motion removal


; ENERGY MINIMIZATION OPTIONS
emtol= 10 ; Force tolerance
emstep   = 0.01 ; initial step-size
niter= 20 ; Max number of iterations in relax-shells
fcstep   = 0 ; Step size (ps^2) for minimization of
flexible constraints
nstcgsteep   = 1000 ; Frequency of steepest descents steps when
doing CG
nbfgscorr= 10


; OUTPUT CONTROL OPTIONS
nstxout  = 5 ; Writing full precision coordinates every
ns
nstvout  = 5 ; Writing velocities every nanosecond
nstfout  = 0 ; Not writing forces
nstlog   = 5000  ; Writing to the log file every step 10ps
nstcalcenergy= 100
nstenergy= 5000  ; Writing out energy information every
step 10ps
nstxtcout= 2500  ; Writing coordinates every 5 ps
xtc-precision= 1000
xtc-grps = Protein_Carb  Water_and_Ions ; subset of
atoms for the .xtc file.
energygrps   = Protein_Carb  Water_and_Ions ; Selection of
energy groups


; NEIGHBORSEARCHING PARAMETERS
nstlist  = 10 ; nblist update frequency-
ns-type  = Grid ; ns algorithm (simple or grid)
pbc  = xyz ; Periodic boundary conditions: xyz, no,
xy
periodic-molecules   = no
rlist= 1.4 ;  nblist cut-off
rlistlong= -1 ; long-range cut-off for switched
potentials


; OPTIONS FOR ELECTROSTATICS
coulombtype  = PME ; Method for doing electrostatics
rcoulomb = 1.4 ;
epsilon-r= 1 ; Relative dielectric constant for the
medium
pme_order= 10;


; OPTIONS FOR VDW
vdw-type = Cut-off  ; Method for doing Van der Waals
rvdw-switch  = 0 ; cut-off lengths
rvdw = 1.4 ;
DispCorr = EnerPres; Apply long range dispersion
corrections for Energy and Pressure
table-extension  = 1; Extension of the potential lookup tables
beyond the cut-off
fourierspacing   = 0.08;  Spacing for the PME/PPPM FFT grid


; GENERALIZED BORN ELECTROSTATICS
gb-algorithm = Still; Algorithm for calculating Born radii
nstgbradii   = 1; Frequency of calculating the Born radii
inside rlist
rgbradii = 1; Cutoff for Born radii calculation
gb-epsilon-solvent   = 80; Dielectric coefficient of the implicit
solvent
gb-saltconc  = 0; Salt concentration in M for Generalized
Born models


; Scaling factors used in the OBC GB model. Default values are OBC(II)
gb-obc-alpha = 1
gb-obc-beta  = 0.8
gb-obc-gamma = 4.85
gb-dielectric-offset = 0.009
sa-algorithm = Ace-approximation
sa-surface-tension   = -1; Surface tension (kJ/mol/nm^2) for the SA
(nonpolar surface) part of GBSA - default -1



; Temperature coupling
tcoupl = nose-hoover
nsttcouple   = 10 ;
nh-chain-length  = 10
tc-grps  = Protein_Carb  Water_and_Ions ; Groups to
couple separately
tau-t= 1010; Time constant (ps)-
ref-t  = 270.0 270.0; reference temperature (K)


; pressure coupling
pcoupl   = no  ;-


; GENERATE VELOCITIES FOR STARTUP RUN
gen-vel  = no
gen-temp  = 270.0
gen-seed = 173529


; OPTIONS FOR BONDS
continuation = yes ;  constrain the start configuration

constraints  = all-bonds
constraint-algorithm = lincs ; Type of constraint algorithm-
lincs-order  = 4
lincs-iter   = 1
lincs-warnangle  = 30


Thank you for your help.

Nisha



On Thu, Jun 30, 2016 at 6:21 PM, Justin Lemkul  wrote:

>
>
> On 6/30/16 8:46 AM, NISHA Prakash wrote:
>
>> Dear all,
>>
>> I have conducted a 10ns REMD simulation for a protein ligand complex with
>> the temperature range - 270 

Re: [gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[Justin Lemkul]

On 6/30/16 8:46 AM, NISHA Prakash wrote:

Dear all,

I have conducted a 10ns REMD simulation for a protein ligand complex with
the temperature range - 270 to 350 K, however the temperature distribution
plot of the replicas show that the sampling has occurred at higher
temperatures as well that is beyond 350K -
Below is an excerpt from the temperature xvg file


@title "Gromacs Energies"
@xaxis  label "Time (ps)"
@yaxis  label "(K)"
@TYPE xy
@ view 0.15, 0.15, 0.75, 0.85
@ legend on
@ legend box on
@ legend loctype view
@ legend 0.78, 0.8
@ legend length 2
@ s0 legend "Temperature"
0.000.00
   10.00  350.997864
   20.00  353.618927
   30.00  350.068481
   40.00  353.921753
   50.00  359.485565
   60.00  353.463654
   70.00  352.015778
   80.00  350.657898
   90.00  351.927155
  100.00  354.539429
  110.00  354.287720
  120.00  349.436096
  130.00  352.960541
  140.00  351.631317
  150.00  354.217407
  160.00  350.185852
  170.00  350.294434
  180.00  350.980194
  190.00  350.914429
   
   
 470.00  349.224060
  480.00  350.819458
  490.00  348.541748
  500.00  350.393127
  510.00  398.775208
  520.00  444.802856
  530.00  470.899323
  540.00  466.652740
  550.00  465.600677
  560.00  469.22
  570.00  470.548370
  580.00  470.011566
  590.00  470.643951
  600.00  472.433197
  610.00  470.451172
  620.00  469.991699
  630.00  469.073090
  640.00  467.259521
  650.00  464.561798
  660.00  468.416901
  670.00  468.754913
  680.00  469.259613
  690.00  467.641144
  700.00  468.542328


Temperature coupling was done using Nose hoover algorithm.

Does this imply the sampling is wrong or insufficent?
Any help / suggestion is appreciated.



How large is your system, and what is it?  What were your (full) .mdp settings? 
The fact that your temperature started at 0 K and ramped up suggests that you 
did not equilibrate prior to the run, did not generate appropriate velocities, 
or did not continue properly.  The sudden jump in temperature later suggests 
instability, and could be due to incorrect settings.  N-H allows for large 
oscillations, but I wouldn't expect a stable system to that degree.


-Justin

--
==

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalem...@outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==
--
Gromacs Users mailing list

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

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a 
mail to gmx-users-requ...@gromacs.org.

[gmx-users] REMD - replicas sampling in temperatures beyond the assigned range

[NISHA Prakash]

Dear all,

I have conducted a 10ns REMD simulation for a protein ligand complex with
the temperature range - 270 to 350 K, however the temperature distribution
plot of the replicas show that the sampling has occurred at higher
temperatures as well that is beyond 350K -
Below is an excerpt from the temperature xvg file


@title "Gromacs Energies"
@xaxis  label "Time (ps)"
@yaxis  label "(K)"
@TYPE xy
@ view 0.15, 0.15, 0.75, 0.85
@ legend on
@ legend box on
@ legend loctype view
@ legend 0.78, 0.8
@ legend length 2
@ s0 legend "Temperature"
0.000.00
   10.00  350.997864
   20.00  353.618927
   30.00  350.068481
   40.00  353.921753
   50.00  359.485565
   60.00  353.463654
   70.00  352.015778
   80.00  350.657898
   90.00  351.927155
  100.00  354.539429
  110.00  354.287720
  120.00  349.436096
  130.00  352.960541
  140.00  351.631317
  150.00  354.217407
  160.00  350.185852
  170.00  350.294434
  180.00  350.980194
  190.00  350.914429
   
   
 470.00  349.224060
  480.00  350.819458
  490.00  348.541748
  500.00  350.393127
  510.00  398.775208
  520.00  444.802856
  530.00  470.899323
  540.00  466.652740
  550.00  465.600677
  560.00  469.22
  570.00  470.548370
  580.00  470.011566
  590.00  470.643951
  600.00  472.433197
  610.00  470.451172
  620.00  469.991699
  630.00  469.073090
  640.00  467.259521
  650.00  464.561798
  660.00  468.416901
  670.00  468.754913
  680.00  469.259613
  690.00  467.641144
  700.00  468.542328


Temperature coupling was done using Nose hoover algorithm.

Does this imply the sampling is wrong or insufficent?
Any help / suggestion is appreciated.

Thanking you in anticipation.

Nisha
-- 
Gromacs Users mailing list

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

* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a 
mail to gmx-users-requ...@gromacs.org.