600 ps of MD is not enough to conclude much of anything. Rg of a folded protein is not usually used (it's generally useful for unfolded/disordered systems, polymers, etc) and may be affected by disordered tails that have no functional significance. A difference of 0.01 nm is meaningless.

600 ps of MD is not enough to conclude much of anything. Rg of a folded protein is not usually used (it's generally useful for unfolded/disordered systems, polymers, etc) and may be affected by disordered tails that have no functional significance. A difference of 0.01 nm is meaningless.

Dear Dr. Lemkul.

I didn't notice that difference in Rg in totall but I thought this difference over a time range means something. For that, I have used a statistical test to find any significant mean differences between receptor and receptor/ligand RMSD and Rg outputs. I have attached RMSD and Rg changes in .csv sheet and configuration file as well. As showed in attached MS-Excel file, RMSD and or Rg changes in some range time is more than 0.01nm, and if performing t-test be rational, it was found to be significant in both RMSD and Rg between receptor and its complex with ligand. However, to perform MD in reliable form, recommend me to how run MDS please.

The RMSD of the whole folded protein, does not indicate the interaction of the ligand, if the protein has a large molecular weight, meaning a large number of atoms, the RMSD of the interaction site will be masked by the fluctuation of the whole protein. You must perform relaxation runs of the protein, an annealing run, then perform at least a 1 ns production run with the ligand using a docked and low energy pose. I recomend explicit waters. Once the production run is performed, analyze the beta factors of the ligand region, the distance and energies between the posible interacting groups of the ligand in the binding pocket. If the ligand does not remain in the binding pocket, obviously it is not stable, perhaps will be necesary to analize another docked pose. 

Dr. Hernandez

By RMSD, I didn't mean protein size changing but stability. I have used the ligand posed in the receptor with the lowest affinity, which has obtained by AutoDock Vina. H-bond, on the other hand was in minimum number at the end of MDS in water solvation, equilibrate system. The files are attached. The ligand has been attached in the receptor through the simulation. There is another question, if an inhibitor stand in its binding site within the target, does it change the receptor stability? If it dose, what exactly the RMSD changes obtained here (in the first figure) suggest?

Some ligands modulate the stability of the protein, others don't. I reiterate that a few hundred ps of simulation is inadequate to address such issues, and RMSD is the least sensitive metric of protein stability, because it doesn't really (on its own) give you insight into what's going on. You should be looking at native contacts, intra-protein hydrogen bonds, secondary structure evolution, etc. But all of these dynamics change on the scale of potentially hundreds of nanoseconds or microseconds.

Thank you dear dr.Lemkul, I understand now.

If it is needed to do it once again, is 100ns is enough? For which, I changed the md file parameters as follow:

; Run parameters

50000000 steps, 100000.0 ps.

is it right?

There is never any way to know a prior if any amount of time is "enough" for a simulation. You can form expectations based on the time scale of the relevant dynamics, but until you actually do the simulation, observe its behavior, and perform analysis including convergence checks, you can't say if anything is "enough."

I will remember that. There is another question. Is it rational to calculate ligand (which contained less atom numbers and may take much less time) dynamics parameters like RMSD etc and substitute them from those in ligand/receptor complex to obtain the receptor alone parameters?

May I ask  how to check out convergence, I mean what MDS parameters is good to be checked. Different studies have used different parameters like RMSD, RMSf, and Rg etc. Actually, I wan to analysis a several ligands binding within a receptor and its mutants. Therefore, extending mdrun may not be even necessary. 10E5ps extension is calculated by my system to lasts a week and it may produce over 10gb outputs just for a single simulation.

Simulating the ligand alone tells you nothing about the dynamics of the ligand when bound or about the receptor. If there is suspected conformational restriction and you're looking at, e.g. entropic contributions to binding, then you have to do both simulations (bound and unbound).

Convergence means that any property of interest is unchanging with time (statistically speaking).  You can't measure any one quantity and necessarily say the simulation is converged. The properties that you are interested in, the ones that answer your posed scientific questions, have to be stable for sufficient time (based on actual physical reality) to conclude that the dynamics are relevant.

Brilliant. So by MD simulation of bound and unbounded structures, each one of structures should be in less fluctuations and stay stable, yet, there must be differences between both bound and unbound structures properties. Suggesting ligand-induced conformational restriction. If one cannot tell a simulation is converged necessarily, how can one say the structures is unchanged?

I have extended mdrun for almost 4ns and still running. RMSD changes of previous and current simulation is attached. Dose this mean somthing?