GROMACS version: 2019
GROMACS modification: No
Hello All,
I am performing SMD-Umbrella Sampling for my system. I used the final frame of converged MD simulation, as a starting structure for SMD.
I got the PMF profile as follows.
My question is,
When considering the final converged structure (assuming our system reached minima in potential energy surface) as starting structure, the PMF curve should begin at this bound state and reach the unbound state. What is the reason behind the dip in this curve? What are the things, we need to understand the scientific or theoretical justification for this state, as it is almost 1.5 kcal/mol less than the starting structure?
Kindly share the knowledge. Thanks in advance.
There is nothing special bout the final frame in a simulation. Let us assume that the initial unrestrained simulation had relaxed sufficiently from starting conditions such that it was sampling from the equilibrium distribution. That only means that configurations are sampled with P=exp(-BE)/Z. The relation between the final frame and a representative frame (if there is one) depends on the partition function (Z). Worst case scenario, P(i)=1/N for all i states. Also bad would be an evenly split bimodal distribution, and you are then guaranteed to only sample pulling out of one basin. Even if you avoid these scenarios, the simulation will likely bounce around the local minimum, so if you started not exactly at the minimum then you might expect to see such a dip as you pass through the minimum on your way out of the basin. One approach would be for you to take your initial unrestrained simulation and plot a PDF along xi, then randomly pick a structure corresponding to the most likely value of xi. This neglects orthogonal DOFs, which you might also be able to interrogate with domain-specific sytem knowledge.
1 Like
