Many 'special' types of MD simulations exist to explore conformational space faster, rather than staying in the same potential energy minimum all the time. I am not familiar with gromacs, but I am sure it will have some of the techniques implemented.

1) Replica exchange (http://dx.doi.org/10.1016/S0009-2614(99)01123-9) runs multiple simulations at once at different temperatures. The aim is for the high-temperature simulations to explore more conformational space whereas the low-temperature simulations still give you the correct boltzmann statistics.

2) If you know what transition you want to see, can can look into steered MD (http://www.ks.uiuc.edu/Research/smd_imd/). The idea is to force the transition you want to see and back-calculate the real potential energy that is involved in this transition.

3) Meta-dynamics (http://www.ks.uiuc.edu/Research/smd_imd/) is another technique which allows you to explore conformational space faster. As the simulation goes on, conformations that are already sampled will be penalised, so new conformational space is sampled.

4) ... I'm quite sure more things exist

If you want to get more efficient sampling, maybe you can try Replica exchange MD or meta simulation.

Enhanced sampling methods such metadynamics and accelerated molecular dynamics are some of the recently developed methods out of many possible methods available to sample rare conformational transition within limited simulation time. But again it depends strongly on your system and the properties you are interested to see.

It can really depends on the kind of protein you are studying. If you don't have any idea on the possible conformational changes, and if it's not a membrane protein, I think the best way should be parallel tempering. Although a bit computationally expensive (since you should perform several replicas at different temperatures) it should give you a suitable sampling of the conformational space. I suggest you to look at the following paper for instance: Earl and Deem, Phys. Chem. Chem. Phys. 7 (2005) pp. 3910-3916. (and the references therein).

I guess for accelerating an MD simulation you have only two options (MD is intrinsically a time-consuming process):

1- Elevating computational power by using more powerful CPU

2- Decreasing the time of simulation

Although there is some other option, which can decrease the time of simulation, but they can affect your final results too.

If your final trajectory file is too crowded you can increase the time step for taking snapshot of the system.

Many 'special' types of MD simulations exist to explore conformational space faster, rather than staying in the same potential energy minimum all the time. I am not familiar with gromacs, but I am sure it will have some of the techniques implemented.

1) Replica exchange (http://dx.doi.org/10.1016/S0009-2614(99)01123-9) runs multiple simulations at once at different temperatures. The aim is for the high-temperature simulations to explore more conformational space whereas the low-temperature simulations still give you the correct boltzmann statistics.

2) If you know what transition you want to see, can can look into steered MD (http://www.ks.uiuc.edu/Research/smd_imd/). The idea is to force the transition you want to see and back-calculate the real potential energy that is involved in this transition.

3) Meta-dynamics (http://www.ks.uiuc.edu/Research/smd_imd/) is another technique which allows you to explore conformational space faster. As the simulation goes on, conformations that are already sampled will be penalised, so new conformational space is sampled.

4) ... I'm quite sure more things exist

In addition to all previous suggestions, you might want to check your force-field parameters. Some versions of some force fields will stabilize the structure more than others.

As the other said, it really depends on the phenomena that you want to simulate. Here you find a good review on the various methods and in particular on metadynamics http://iopscience.iop.org/0034-4885/71/12/126601

Expecially if you are interested in using several collective variables in metadynamics as in protein folding problems, I suggest you to use Bias-Exchange metadynamics (http://people.sissa.it/~laio/Research/Res_BE.php , http://www.pnas.org/content/110/17/6817.short )

To use gromacs, PLUMED is a great plug-in for most of the enhance sampling technique (http://www.plumed-code.org/)

Hi Zalikha,

Why don't you try the C-alpha model builder as implemented by Paul Gerber in the very versatile molecular modelling suite Moloc?

Free for Universities:

http://www.moloc.ch

Wish you a fine success,

Christian

There are lots of method. More expensive but powerful methods

1. Transition path sampling

2. meta dynamics ( implemented in NAMD)

3. Replica exchange (with umbrella sampling or temperature)

4. milestoning

5. nudged elastic band

6. Umbrella Sampling

Easy methods already implemented in softwares

1. Adaptive biasing force (ABF)

2. Accelerated MD

3. Targeted MD

etc

Hi,

I am not sure i understand your question. You say you are looking for a " rare conformation". What do you mean by this? Are you truly just looking for a single conformation, which basically means a state of protein at say one point in time. If that's truly the case then i personally don't think MD is the correct approach to solve this problem. With MD you have a limited scope of effectively sampling, you are limited by your starting conformation and then each subsequent conformation that appears is as a consequence of the previous one, hence the trajectory. If you mean what i think you do, that you are actually looking for as you say " rare conformation" then MD is not the way, you should choose a Monte Carlo based method. With that your simulation would be independent of the starting conformation and each conformation (that results) will have an equal probability of being visited increasing your chances significantly of obtaining a " rare conformation" as compared to MD. However with Monte Carlo you will not be able to see the way in which the " rare conformation" was reached but that can be guessed to some degree ...

Hope this helps,

Best,

/A

Another feasible option could be accelerated molecular dynamics, a method published by McCammon and Ross Walker in 2012, which essentially let you sample configurations observerd in the milisecond time-scale.

Other alternatives, as pointed above, are metadynamics, replica exchange, if you know something about the target configuration, then use target or steered molecular dynamics.

If your protein configuration doesn't take care about the time, then you can also use either monte-carlo approaches or coarse grained methods, such normal mode analysis.

But at the end is a more phylosophical question, because if you don't really know what to expect as outcome (how to define a strange protein configuration or conformation based on a initial protein structure?), is difficult to recommend you any choice.

Good luck!!

Dear all: My aim is to screen if there is any other conformation especially at the region near the binding pocket. I see none so far - the conformation may possibly stuck in local minima as what Pieter pointed in his comment.

I did run the simulation at bigger timestep, but it crashed. I thank all of you for your wise comments and I am considering all suggestions given above :)

Dear Stéphane: Thanks for making an awareness about that. It was actually my attempt to speed up the simulation, but it turned out to blow up - making me realized that this could not be applied.

Exploration about the enhanced sampling approaches is currently going on. Wish me luck! Will update if anything.