There may be too many parameter files varying according to your requirement. Clearly define your purpose here.

There may be too many parameter files varying according to your requirement. Clearly define your purpose here.

I think You may look for some tutorials etc. that are similar to Your system, and try to use same .mdp byt it probably won't be perfect. As Rajnikant wrote, i worry there is nothing like universal "best parameters" for everything.

In the *.mdp file you tell GROMACS what kind of simulation you wanna do.

(-> http://manual.gromacs.org/current/online/mdp.html )

This includes the ensemble type (e.g. NVE, NVT, NPT,...) and many other fundamental definition.

So, the question about 'best molecular dynamics parameters' makes absolutely no sense!!

Before you try to parametrize GROMACS, you should first understand what you are doing! Start with a text book about MD!

Why I ask this question because there is disturbance between my aim and OUTPUT of GROMACS energy minimized structure

example I model a structure through modeller and submitted to Ramachandran plot it's given 89% allowed region, I did energy minimization of same structure through GROMACS,now output structure from GROMACS submitted to Ramachandran plot it's given 84% allowed region

It is very possible. Energy minimization might for example fall to local energy minimum. You can try simulated annealing to overcome it. But it is just the simpliest explanation. Also You can try Conjugate gradient method instead of steepest descent. It also helpssometimes.

Also having less allowed residues after Gromacs processing (i understand You minimalized Modeller output?) may come from some unstructurized loops eg.that were predicted by Modeller basing on crystalized structures.

Sorry Shameer, I don't intend to be rude. Now your question is much clearer.

I would like to note, that it might be worth to check if the potential model you utilize is capable to give good prediction for your molecule of interest.

And Stéphane is right, may be the influence of the solvent is crucial here for the structure.

Frank Römer why sorry for us you guide me to display properly the problem

Which force field is best for minimizing of protein AMBER or CHARM

How much diameter to be define for a BOX

After this step in my results "No velocities are found "

Shameer here i would like to add for your box dimension question that Box use in MD is for defining the boundaries of a system. Inside which only GROMACS creates environment for real MD. So varying with every case you can keep your box dimension so that it covers your system basic molecule entirely and a bit more. i.e. Basic molecule dimension + 1-2 Angstron.

The comments from Stéphane about box dimensions and PBC is really an important issue.

About the potential model/framework:

One can not directly compare AMBER with CHARMM (or others) in general. If I would like to find a suitable potential framework for a given protein (or in general molecule), I would check if there are any reliable publications regarding that kind of protein (or molecule) class.

So, if you think you could argument in a paper that your choice of framework is reasonable for your problem, it might be a good start.

@Stephane. Thank you for clearing fact better. My only concern was to clear Mr. Shameer doubt about Box in MD.

Stéphane Abel you clarified my doubt regarding the box based on the protein size and free space for simulation process must be in proper dimension

MMM... to me, the discussion have moved away from the starting problem:

Why the minimized structure does loose quality (as judged form Ramachandra's plot) when subjected to MD relaxation?

There are many faces to this problem:

A) MD simulations rely on force-field, which are empirical constructs attempting to approximate the energy profile of a molecular system. In the very end, they are flawed because all quantum chemistry contributions are ignored. See for instance:

Faver JC , Benson ML , He X , Roberts BP , Wang B , Marshall MS , Sherrill CD , & Merz KM Jr. (2011) The Energy Computation Paradox and ab initio Protein Folding. PLoS ONE 6(4):e18868 (http://dx.doi.org/10.1371/journal.pone.0018868).

B) Since force-fields are empirical and have a bias, they all have a tendency to deviate from the canonical values we observe in natural structures, each one with its own tendency. The one that gets closer to the PDB structures is the ROSETTA forcefield

(Martínez-Castilla LP & Rodríguez-Sotres R (2010) A score of the ability of a three-dimensional protein model to retrieve its own sequence as a quantitative measure of its quality and appropriateness. PLoS One 5(9):e12483. ,

http://dx.doi.org/10.1371/journal.pone.0012483

;

Kuhlman B , Dantas G , Ireton GC , Varani G , Stoddard BL , & Baker D (2003) Design of a novel globular protein fold with atomic-level accuracy. Science 302(5649):1364--1368. ,

http://dx.doi.org/10.1371/journal.pone.0012483).

In the first of these two papers you will find information regarding the differences in Ramachandra`s plot for NMR and X-ray solved protein structures. Therefore, even experimental methods do produce different "equilibrium" structures.

SO if you want to have a model closer to what is observed in the PDB your best choice would be relaxing it with ROSETTA (try the ROSETTA fast-relax protocol). However, even ROSETTA deviates from the classic PDB structures (Chavelas Adame EA , Hernández-Domínguez EE , Gaytán-Mondrangón S , Rosales León L , Valencia-Turcotte L , & Rodríguez-Sotres R (2011) A Hitchhiker's Guide to the modeling of the three-Dimensional structure of proteins. International Color Biotechnology Journal 1(1):26-35.

http://www.cicy.mx/Sitios/Journal/pdf/2011/October/HighDefinition/IBCJ_2011_1_1_26S.pdf

;

Rosales-León L , Hernández-Domínguez EE , Gaytán-Mondragón S , & Rodríguez-Sotres R (2012) Metal binding sites in plant soluble inorganic pyrophosphatases. An example of the use of ROSETTA design and hidden Markov models to guide the homology modeling of proteins. Journal of the Mexican Chemical Society 56(1):23-31.

http://www.jmcs.org.mx/PDFS/V56/1/05.-%20Rosales-Leon.pdf

).

The most important problem of structure prediction is not getting the structure, for there are many servers to make a prediction, but to measure the quality and the biological appropriateness of the prediction (see Martínez-Castilla & Rodríguez-Sotres, 2010 for a discussion of this aspect, and check Chavelas-Adame at al., 2011 for help on how to produce and score a protein 3D model).

As implied in my previous comment, force-fields do have significant differences. Modeller uses CHARMM, and Gromos, can use several force-fields (Gromos in many flavors, several Amber versions, Encad and some others) but by default uses GROMOS-G53a6, in the later versions. You can however use CHARMM in Gromacs if you decide to do so, because this force-field has already been ported into GROMACS. Look at the GROMCAS documentation for details.

Best wishes,

Rogelio

Thank Rogelio Rodríguez-Sotres for your valuable and detailed answer on PDB structures and force filed to be use, you provided valuable papers link