I suggest a slight modification on what Martin has suggested. Running many 10-ns MD simulations of a membrane is not going to be very useful; the lipids will still not have equilibrated in that time. Several replicates of 50-100 ns would be more appropriate, e.g. 10 x 100 ns would be better sampling than 1 x 1 microsecond, unless the dynamics of the embedded protein also must be observed over longer time.

OMG!!

You have performed MD for 100 ns, its great. As far as I can see, at the end, the trajectory doesnot stabilise. But keeping view of the fact that upto at least 60ns, the trajectory was stable, its preddy fine. I really not able to understand why you want to extend beyond 60 ns?

I will be happy to know your aim behind this longer simulation run?

Show the scale of the vertical axis.

---

Calculate average RMSD separately for the first and second half of the trajectory. Set an arbitrary, reasonable threshold; if the two average RMSD values are below the threshold, consider the simulation, from RMSD perspective, as converged.

Molecular dynamics (MD) simulation is a controlled explosion. Running many short MD simulations (few nanoseconds) starting from different initial conditions (e.g., different initial structure or different initial random velocities) is usually better than running one long MD simulation, unless one wants to study events that cannot begin and complete within a short simulation time.

If you decide to extend the simulation, do it by, at least, one order of magnitude, i.e., in your case: 1.0 microsecond. But you will probably get much better conformational coverage, and faster, by running one hundred 10-ns long MD simulations.

---

Plot the cumulative average RMSD. It's quite flat at the end, isn't it? The more data you collect, the more it takes to change the average. That's why extending the simulation by less than 1-2 orders of magnitude will probably produce nothing more useful than what you already have.

I suggest a slight modification on what Martin has suggested. Running many 10-ns MD simulations of a membrane is not going to be very useful; the lipids will still not have equilibrated in that time. Several replicates of 50-100 ns would be more appropriate, e.g. 10 x 100 ns would be better sampling than 1 x 1 microsecond, unless the dynamics of the embedded protein also must be observed over longer time.

I also think that I will have to extend my simulation, maybe by 100 ns more which is very very computationally expensive for me, but I will give it a try for one last time. But, I still haven't been able to figure it out why my rmsd is jumping up and down from 70 to 90 ns , though RMSD is pretty flat till 60 ns. Is it due to loop shift conformation or something, but visually I can see no change in conformation.

Rather than making movies in VMD, I have attached here with the screencast video I made of the simulation. I've also attached here with corresponding RMSD plot I created with VMD. I can't see any significant structural/conformational change in dimer from Frame No. 3000 to Frame No. 4500, corresponding to its drastic RMSD change in the plot. For the sake of proper visualization, I've hidden lipids and water from the view.

https://drive.google.com/file/d/155ZRUcMRiJkLeAdWlAscY6q5exED0J9L/view?usp=sharing

Hi Roshan Shrestha you have mentioned "But, I still haven't been able to figure it out why my rmsd is jumping up and down from 70 to 90 ns , though RMSD is pretty flat till 60 ns. Is it due to loop shift conformation or something, but visually I can see no change in conformation".

For that, check SSE (secondary structure elements) distribution by residue index throughout the protein structure. The plot summarizes the SSE composition for each trajectory frame over the course of the simulation.

Hi,

Instead of increasing the dynamics time, I suggest increasing the previous step of equilibration (e.g. constant temperature). the plot would be better if the dynamics start flat from beginning.