GROMACS version: 2023.5
GROMACS modification: Yes/No
Here post your question
Hello Guys, I am running umbrella sampling I found for some particular frame it is very hard to achieve overlappinng with neighborinng frames, even if I used continued frames from extracted from the pulling simulation. I then tried the trick of increasing the force constant, but it either showed double peak or peak not overlapped well. For the double peak one, it overlapped well with neighbouring peaks. My understandinng is that the double peak is still a biased distribution. So is this frame with double peak still usable?
how is your pmf profile?
it looks fine, but depends on what simulation system is this.
Hi there. It is a protein complex (protein data bank accession number 2qc1)
Are you pulling the molecule away ?
Yes. The pulling pulled the binding partner 5 nm apart in all.
Do you expect a binding free energy of ~20 kcal/mol? What force constant are you using in the umbrella sampling? The slope of the PMF is quite steep, so you might need a quite high force constant. Are the starting positions in the umbrella sampling evenly distributed, i.e. are the frames extracted from the pulling simulation evenly distributed?
If you are pulling away, the binding energy should reduce over time, the graph is reversed? can you show the last frame of your pulling, you need to pull and make sure full dissociation of the molecule.
Hi MagnusL, Yes. The binding free energy is around 17 kcal/mol. I started with a force constant of 1000 kJ mol^-1 nm^-2 and do you suggest higher values? I did tried higher values for some of the frames and it did help but for some particular ones it didn’t even if I increased the value to 2500. I tried the 5 nm pulling distance and found the pmf curve increased steeply from 2.7 to 3.5 nm. Do you suggest that I use denser frames and increase the force constant? I copied the pulling and umbrella sampling md parameters from Justin’s beta-amyloid peptides umbrella sampling’s tutorial and the force constant of 1000 kJ mol^-1 nm^-2 was tried. The distance is 2.869, 3.062, 3.150, 3.256 for the initial four frames. The third frame was added afterwards when I found the overlapping was not good.
Hi scinikhil,
Thank you for your reply, but I don’t understand why you think the graph should be reversed. The pmf curve should be similar as the one from Justin’s umbrella sampling tutorial. The binding energy is the difference between the pmf curve top and bottom.
You are right, i was thinking in opposite way of pulling :)
Thank you for your confirmation!! :)
Good that binding free energy is close to the expected value at least. Then it might already be OK.
I would think 2500 kJ mol^-1 nm^-2 and more umbrellas (3 times as many?) would be slightly better, but definitely more computationally expensive. Since you said it didn’t help with a higher force constant it might not be worth it. I’d definitely vote for another umbrella at 3.3 nm at least. But that won’t make a large difference.
Hi MagnusL, thank you for your suggestions!!