AMBER Archive (2007)

Subject: Re: AMBER: convergence in classical MD

From: Carlos Simmerling (carlos.simmerling_at_gmail.com)
Date: Mon Jul 16 2007 - 06:37:09 CDT

I think you should generate independent runs from different initial structures
and continue them until the differences between ensemble averages from the
runs are small enough to provide acceptable error bars on your data.

On 7/16/07, Seongeun Yang <seongeun_at_korea.ac.kr> wrote:
> Hi all,
>
> I have a question on the convergence problem in classical molecular dynamics.
> The ff99SB was used in generating topology files.
> The system of interest is a alanine-based peptide solvated with TIP3P water molecules.
> Whether I used NPT condition through all simulations or
> heated the system under NPT for 50 ps stepwisely and then equilibrated it under NVT condition,
> the average helicity and Ca rmsd values do not converge until 10 ns.
> Each trajectory is generated using different initial velocities (i.e. different ig values)
> and the average was over 10 trajectories.
>
> For reference, the input I used in NVT simulations is;
> -----------------------------------
> &cntrl
> imin = 0, ntx = 5,
> irest = 1, iwrap = 1,
> ntpr = 500, ntwe = 500,
> ntwr = 500, ntwx = 500,
> ntave = 50000,
> ntf = 2, ntc = 2, ntb = 1,
> cut = 8.0, nsnb = 10,
> nstlim = 500000, dt = 0.002,
> tempi = 300.0, temp0 = 300.0,
> ntt = 1, tautp = 1.0,
> &end
> ------------------------------------
>
> For NPT simulations, the only difference was with ntb=2 and an additional line like
> ------------------------------------
> ntp = 1, pres0 = 1.0, taup = 1.0,
> ------------------------------------
>
> Should I do longer than 10 ns simulations until I see convergence?
> Or something missing or wrong?
>
> Please let me know some useful tips for this problem.
> Thanks.
>
> Seongeun
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu