RMSD graph after protein protein md simulation

User discussions
, , , ,
1

GROMACS version:
GROMACS modification: Yes/No
Here post your question
I did protein-protein MD simulation and did equilibration for 500ns. First for 200ns then extended till 500ns. After pbc mol, centering,in RMSD i see the graph is fluctuating, till 250ns apprx. the rmsd is stable then sudden jumps in 300 ns and stable till 500ns. I checked the trajectory motions and I observed, till 200-250ns the protein stays in complex but after that it drifts apart and stay unbound till 500ns. Is this normal? the production run trajectory motions also showed unbound state over 500ns run. is this normal? or i have to apply gentle force to the system to keep it together? Whats the point of the complex if it drifts apart and stays independent?

rmsd200
rmsd200792×612 35.9 KB

I am unable to understand what to do? should i go for fresh run or any other possible solution? I need suggestions on this.

rmsdNPT500
rmsdNPT500792×612 32.9 KB

I attached the initial 200ns quilibration run and then extended equilibration run. also the rmsd production rmsd.

rmsd600nsPDRUN
rmsd600nsPDRUN792×612 38.2 KB

2

I guess this is an RMSD plot of the backbone of the whole structure, so both proteins together?

First, I wouldn’t say that for the first 250ns the system is stable. Already after 15ns or so you have large fluctuations up to ~1.3 nm, and same towards the end of the 250ns, which in principle are quite large numbers. The fluctuations are telling you that the system is re-organizing itself a lot, and most likely detaching and reorienting, although it’s hard to state anything as the RMSD is an extremely degenerate metric. One of the things that should be done here is check the trajectory, and as you said you see the system detaching and flying apart, so you have a reason for these numbers: the dimer doesn’t stick together.

Is this normal? the production run trajectory motions also showed unbound state over 500ns run. is this normal? or i have to apply gentle force to the system to keep it together? Whats the point of the complex if it drifts apart and stays independent?

I am unable to understand what to do? should i go for fresh run or any other possible solution? I need suggestions on this.

These are questions for you, as your simulations will be based on your scientific inquiry. Some proteins interact and stick together, some other don’t, there is no normal or expected behavior, and there is no general solution. Applying a force to keep them together will solve the problem of them detaching, but you need a justification. Enforcing, for example, a wrong interface will give just garbage as output results. Personally, I would check

i) if the force field is suitable for this complex. Am I using a ff that is notoriously bad at reproducing the quantities I’m looking for?

ii) What has been done previously in literature? Did someone simulate my complex? What problems do they have?

iii) my starting structure. Is this a crystal? Is this a complete structure or am I neglecting part of the proteins that can stabilize the interactions, like a membrane domain or an interacting loop?

iv) my equilibration protocol. Am I being too brutal? Maybe a slow warm up of the system with hard position restraints is needed.

v) run more than one replica. Nowadays, any result should not be presented as a single replica, but should be repeated a few times, especially for unbiased sims.

3

Thank you for your response and clarification on this matter. Yes, the rmsd is of a protein-protein complex. The structures were generated using alphafold, although individual Crystal structures were available, we were looking for complex structure which was not previously shown in literature.
Coming to ff, charmm was used and system was relaxed with no force field given at the end. Maybe was it the reason why the complex drifted apart? Also, production run was supposed to be free of any restraints as per my knowledge. So if I have to keep the system together, should it be recommended to keep restraints in both equilibration and production run? Like a 100kj restraints or so? Is that scientifically approved? Or I could assume that my protein complex prefers to stay detached in absence of any force on it. These are the concerns I’m having recently. Trajectory shows clear detachment over 300-600ns. I haven’t made any replica yet as I was not sure if it’s right or not. I was skeptical because I extended these runs instead of running at one go. Should this also be a reason for sudden fluctuations? If my protein protein complex is not bound, I assume there’s no meaning doing drug targetting on this complex.

4

Maybe was it the reason why the complex drifted apart? Also, production run was supposed to be free of any restraints as per my knowledge.

The absence of any particular restraint during the production is, in general, correct, at least in the vast majority of cases of simple unbiased bio sims of protein. There are cases where they are used, but they need to be justified.

So if I have to keep the system together, should it be recommended to keep restraints in both equilibration and production run? Like a 100kj restraints or so? Is that scientifically approved?

There is no real scientifically approved restraint to apply. It does really come down to what you want to do. If you want to target these structures and you expect them to interact in experiments, then in principle they should do so also in simulations. Maybe you can justify in some way a restraint in case you want to maintain a specific surface contact, but again, what is the justification, e.g., do we have a specific interface that we know is present experimentally?

I was skeptical because I extended these runs instead of running at one go. Should this also be a reason for sudden fluctuations?

No don’t worry, properly extending a run is completely harmless.

The structures were generated using alphafold, although individual Crystal structures were available, we were looking for complex structure which was not previously shown in literature.

I don’t know your systems and I am - by far - not a leading expert in protein design and simulations, so take my considerations as reasonable opinions. I think one of the main points here is that if you do not have any crystal structure of the protein interacting then that might be the problem. Was the interface predicted by alphafold? Does it make sense (in terms of electrostatics/hydrophobicity etc)? Do you have any proof that this is the right starting configuration of these interactions? Do you have any experimental proof that these proteins interact and form stable dimer structures of some kind? I would consider these points very carefully before wasting too much time in simulations, as a bad initial guess can easily become a few weeks (if not months) of wasted computational time!

5

Thank you so much for taking time and clarify one by one. I’ll look into it , now that all the issues have been pointed out.