[gmx-users] Technical

Justin A. Lemkul jalemkul at vt.edu
Tue Jun 29 15:58:50 CEST 2010

pawan raghav wrote:
> Dear Justin,
> Thanks for the wonderful suggestions which is very clear and informative 
> but I am confused about some points below
> 1. Actually my protein is human mitochondrial protein so I have 
> used GROMOS 96 53a6 ff. is it correct? but you have mentioned about the 

There is no immediate connection in my mind between a mitochondrial protein and 
GROMOS96 53A6.  How did you jump to this conclusion?

> condition applied, so I don't understand this point because I have a 
> normal protein which bind on the mitochondrial membrane. So according to 

Now here is a major consideration.  Are you modeling the membrane as well, or 
just the protein?  If you are modeling the membrane, the choice of force field 
becomes more complicated.  There are numerous lipid parameter sets - Berger and 
Kukol united-atom sets, OPLS-AA and CHARMM all-atom sets.

> you which conditions should i applied? which preferences you are talking 
> about? Due to have limited configuration of our hardware we are using 
> core2duo single processor with 4 GB RAM. So dear please suggest me the 
> concise requirements for this simulation.

No.  You have to do your own homework.  The most challenging part of any 
simulation is planning it.  You have to do the legwork of reading about the 
protein force fields, and maybe the lipid force fields as well (which you 
haven't mentioned).

> 2. My protein does not depend on the temperature that means it is not 
> thermophilic or mesophilic protein. So for large simulations is it 
> correct to apply REMD technique. I think that ED sampling should be best 
> but don't know which should be taken for reference structure. I have 
> tell you about my system and protein so please suggest me the sampling 
> technique is effective in my protein case.

The type of method you employ is based on what you want to observe.  You said 
you wanted to sample loop dynamics, for which you can either run several 
simulations at the relevant (physiological?) temperature and analyze those 
results to obtain some averages, or you can apply REMD.  I don't know how useful 
ED/PCA would be here, but it is a post-processing analysis technique, so it 
won't help you in sampling.

> 3. I know about the energy conservation but don't know about the 
> simulation stability. So please also tell me is PCA should be performed 
> for simulation stability or something else analysis should be performed.

These two concepts are unrelated.  Stability simply means the model physics 
doesn't break down and your simulation actually runs.

> 4. Why 4-5 fs is more effecient than 3fs?

I mentioned this for several reasons.  First, you collect more data faster.  But 
if you're using lipids (note how this is a major consideration!) then preparing 
these topologies is not trivial.  You basically have to write the topology by 
hand.  Processing a protein with virtual sites is easy with pdb2gmx.  Second, I 
have a minor pet peeve when I see a 3 fs timestep in the literature along with a 
claim like "simulations were conducted for 100 ns using a 3 fs timestep." 
That's mathematically impossible, so I think the reporting is inaccurate.  But 
that's just me :)  And besides, if you're going to extend the timestep, why not 
go all the way and really utilize the performance GROMACS offers?


> -- 
> Pawan


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list