[gmx-users] different output generated by continue and discontinue > simulation

[Justin A. Lemkul]

Hsin-Lin Chiang wrote:
> Hi, Mark
>  > > tc-grps             =  A-chain B-chain drug SOL  NA+
>  > >
>  > >
>  > >
>  > >
>  > >
>  > This is bad - see 
> http://www.gromacs.org/Documentation/Terminology/Thermostats
> Thank you for your website.  I understand now.
>  > > grompp -f md2.mdp -n index.ndx  -p topol.top -c ${n-1}ps.gro  -t 
> ${n-1}ps.trr -o ${n}ps.tpr
>  > >
>  > >
>  > > mdrun -s ${n}ps.tpr -e ${n}ps.edr -o ${n}ps.trr -g ${n}ps.log -c 
> ${n}ps.gro
>  > >
>  >
>  > You are following neither of the approaches recommended here 
> http://www.gromacs.org/Documentation/How-tos/Doing_Restarts
>  >
> Ya, I'm sorry but my colleborator give me that before.
> It is similiar to the commends connect position restrained dynamics and MD.
> You can find grompp read gro and trr of last time frame to make a tpr 
> file of next time frame.
> I thought this cause the time write on each snapshot is 0, but the 
> dynamics is still processing.
> Did I miss something?
> You say the pressure-coupling will be lost in each ps, does it mean that 
> trr file doesn't have this message inside?
> 

Pressure coupling information is stored in the .edr file.  If you don't pass it 
to grompp, you lose it.  The better method is to use a checkpoint (.cpt) file, 
which contains everything necessary to accurately continue the simulation.

>  > > Theses two data generated by different ways are totaly different.
>  > >
>  > > Here I mean different is not on the number buy mean the tendency of 
> figure.
>  > >
>  > >
>  > >
>  > >
>  > >
>  >
>  > See http://www.gromacs.org/Documentation/Terminology/Reproducibility
> Ya, I know this since last time you show me this web in mailing list.
> Thank you again.
> But the different between these two outputs is totaly different.
> I have four systems with different parameter gen_seed.
> First is always unstable and irregular, but second is stable and regular.

It sounds like you have an unstable system, that, under certain favorable 
conditions, seems to work, but in other cases, doesn't.  I'd investigate why the 
first system is failing, since it probably really is a problem in all of them.

http://www.gromacs.org/Documentation/Terminology/Blowing_Up#Diagnosing_an_Unstable_System

>  > > On the contrary, the second method generate a stable movie and 
> fluctuation of interaction energy.
>  > >
>  > > Theotically these two should be the same, right?
>  > >
>  > >
>  > >
>  > >
>  > >
>  >
>  > The second is probably happier about not losing the pressure-coupling 
> information every 1ns. However only much
>  > much longer trajectories should show mutual convergence, and the 
> movie is not a reasonable way to look for it.
>  >
>  > Mark
> Sorry I don't understand you here.
> Do you mean second way seems to more correct?

Not correct, but rather not yet suffering from the same problem.

> And what kind of mutual convergence you mean here?

Systems only converge reliably over long time periods, i.e. in theory, 
independent observations can only be reliably obtained over sufficient time.

-Justin

> Does It mean I can try to find mutual convergence in output of second way?
> I'm sorry for my stupid and poor English.
> 
> Sincerely yours,
> Hsin-Lin
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================