[gmx-users] Re: Simulation in the high temperature conditions

Justin A. Lemkul jalemkul at vt.edu
Tue Apr 10 16:25:56 CEST 2012



James Starlight wrote:
> Justin,
> 
> 
> 10 апреля 2012 г. 16:21 пользователь Justin A. Lemkul <jalemkul at vt.edu 
> <mailto:jalemkul at vt.edu>> написал:
> 
> 
>     I have never used the bd integrator so I cannot offer any help here.
>      I don't see any magical reason why you should expect it to enhance
>     sampling though.
> 
> 
>  I've found in the literature that this integrator could rise the 
> integration-time-step from common 1-2fs used with the MD-integrator.
> 

The timestep is not directly tied to the integrator.  The timestep is decided 
based on the fastest motions in the system.  Using constraints for bonds allows 
an increase in timestep.  There is information about these factors in the 
Gromacs manual, and likely any textbook regarding MD simulation.  A 2-fs 
timestep with the 'md' (leap-frog) integrator when using constraints is quite 
common.

> 
> 
> By the way I've tried to test my system equilibrated in the hight 
> temperature in different conditions. As the consequence when I've 
> removed all posres ( on the last stage of equilibration it's value was 
> 200 for backbone atoms only) from my protein the alpha helixes have been 
> destabilised already after 3ns of such unconstrained simulation ( ref_t= 
> 700k).
> 

That's not surprising.  Most force fields were not designed to reproduce the 
structure of proteins at ambient temperature when exposed to conditions that 
exceed boiling.  Denaturation at this temperature should be expected.

> So I suppose that the presence of the some constrained factor is needed 
> to prevent destabilisation but allow conformation sampling.  I see three 
> alternative ways of my production MD run with enhanced sampling.
> 
> 1- First of all as the most trivial way I'm thinking of using soft 
> posres applied on backbone atoms only with the constained value ( 50-100 
> kj*nm2) corresponded to the energy of the thermal motion.
> 
> 2- Also I've thought about application of the harmonic 
> distance_restraince (instead of posres) on the all backbone atoms exept 
> of flexible loops where this disres must be in the longer range ( e.g up 
> to 15-20A) but I could not realise about usefullness of such aproach 
> because I cant define value for such disres for the atoms in helixes. 

Sure you can.  Use genrestr with a suitable index group.  The use of lots of 
distance restraints precludes the use of DD, so you're limited to using mdrun 
-pd, which is a lot slower and may counteract any perceived benefit in terms of 
simulation time.

> It's known that conformation transitions wich I want to ssample 
> accompanied by the large scale motions in the helical segments ( up to 
> 10-15 A). At the same time such big fluctuations on the other atoms may 
> destabiise regions wich are more rigid. Should I define several sets of 
> restrains for my protein to allow one regions move with bigger 
> amplitudes in comparison with rigid fragments ?
> 

All of this sounds like a large bias on your results and deconvoluting what is 
real versus what you've caused to happen will be a huge headache.

> 3- Finally alternative aproach is the ussage of common temperature ( 
> ref_t=310k) in the production run but using starting velocities obtained 
> from the nvt equilibrration obtained with the hight ( 700k) temperature. 
> Can I enhance sampling using that starting conditions ?
> 

I don't think so.  We've discussed this already.  I suspect your thermostat will 
have a hard time converging and all you've ended up doing is creating a weird, 
undefined ensemble to start your simulations.

> 
> What from this three strategies is the most suitable? IS there any 
> alternative ways for such high-temperature simulation ?
> 

High-temperature MD is not a new concept.  There's lots of literature about it. 
  But if you find yourself having to engineer fancy ways of simply keeping your 
protein stable, it starts getting too fishy to be plausible, IMHO.

-Justin

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

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

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



More information about the gromacs.org_gmx-users mailing list