[gmx-users] changing the velocity in trajectory file

Lutz Maibaum lutz.maibaum at gmail.com
Tue Nov 30 04:45:15 CET 2010


On Nov 29, 2010, at 5:58 PM, sreelakshmi ramesh wrote:
> Basically i have been using a sampling method called FFS to calculate the rate constant for folding  of a protein .The umbrella sampling done on the system shows there is high barrier between the reac an prod states The phase spece regions in ffs method  are divided in to many non -intersecting interfaces along the reaction coordinate.i have started the simulation from reactant region and ended up at a point in between the reactant and product state The transition state is not attained so far.The simulation is stopped between the transition state  and reactant .I have 100 such configuration stored at that state.Now i have to select every config at that state at random  amd it should start with the same velocity when it was stored at that state and continue the simulation to reach the next state.so i am reading the position and velocities of the con-fig from previously saved trajectory of that con-fig.so since randomly if the same con fig s selected twice the trajectory is the same which i don't want.

I assume the FFS you refer to is Forward Flux Sampling. I don't think you can use this method for fully deterministic dynamics, for precisely the reason you describe. You have to introduce some form of stochasticity, but I am not sure if random changes in the velocities at the interface configurations are the best way to do that. You might want to take a look at section 3.2 of http://dx.doi.org/10.1088/0953-8984/21/46/463102

There are, however, legitimate uses of changing the velocities of a configuration that are not complete reassignment from a Maxwell distribution. In Transition Path Sampling, for example, you need to create a trajectory that is "close" to another one for a while, and you usually do that by making a "small" change to a configuration's velocity. The details of how you make this change enter in the acceptance probability of a TPS trial move. I don't think Gromacs provides this functionality out of the box, so you probably have to write your own code to do that.

Hope this helps,

  Lutz




More information about the gromacs.org_gmx-users mailing list