There are many things that could go wrong -- some of which relate to the way you set up your calculation, and some of which are more reflective of the structure upon which you're basing your model. Determining the precise nature of an RMSD drift is thus impossible without a fair bit more information.
I'm going to assume that you applied basic common sense in preparing and equilibrating your preliminary model. Off the top of my head, here is a non-comprehensive list of some other things that can go wrong:
- residue conformations near the hinge points of protein normal modes were not resolved well enough in the starting structure
- the temperature at which your starting structure (and reference) was resolved may be rather low and may not adequately reflect physiologically reasonable conformational variations
- if the starting structure is based on crystallographic characterization, it may be that crystal-forming conditions required some sort of surfactant or other biologically foreign environment.
- if the starting structure is based on crystallographic characterization, the structure may reflect intermolecular interactions (across unit cells) that do not manifest in your simulation
Most of the above may be partially intuited by referring back to the publication reporting the structure you've borrowed for setting up the simulation. If you do so and don't find any plausible sources of discrepancy along the lines of those mentioned above, then drop another note here.
You might also consider communicating details of how you prepared and equilibrated your structure, and whether you are introducing any molecular partners into your system that differ significantly from the experimental environment.
As site editor for MedChemNet, I'm pleased to provide this response from networker Dr Harish Kundaikar:
"The idea of simulation for 10ns seems to be just enough as this is the statistically significant amount of time basically needed to conclude anything from molecular dynamics. However being a medicinal chemist you would need much more simulation time in order to publish in reputed journals.
*Is there any way to find out the reason behind instability or RMSD fluctuation?
- You have not specified your target that you are simulating, you have not even mentioned whether you are simulating a drug-protein or drug-polymer or only protein or polymer system. so answering this question would be difficult. Reasons will vary for either of such systems, you will need to mention some details about this.
*What I can conclude if RMSD is greater than 4 or 4.5 angstrom? If any reference is possible for clearing my doubt/problem so please send me.
- Before starting any simulation you definitely have to relax the system. You may create your own protocol for relaxation or you may use the default protocol provided in Desmond. If you are getting an RMSD of greater than 4 then it means that you have not relaxed the system prior to simulation. Or if you have done the relaxation and still getting such high RMSD then your system has not yet equilibriated, in which case you will need to continue with your simulation further than 10ns.
RMSD higher than 4 is possible for some proteins which are known to disaggregate. One suitable reference I would suggest for good details on simulation using Desmond would be:
“Insights into the Interaction Mechanism of Ligands with Aβ42 Based on Molecular Dynamics Simulations and Mechanics: Implications of Role of Common Binding Site in Drug Design for Alzheimer's Disease.” DOI: 10.1111/cbdd.12555, HS Kundaikar, MS Degani. Chemical biology & drug design 86 (4), 805-812.
Accessible through: http://onlinelibrary.wiley.com/doi/10.1111/cbdd.12555/full
Or through: https://www.researchgate.net/publication/273466298_Insights_into_the_Interaction_Mechanism_of_Ligands_with_A_b_42_Based_on_Molecular_Dynamics_Simulations_and_Mechanics_Implications_of_Role_of_Common_Binding_Site_in_Drug_Design_for_Alzheimer%27s_Disease?ev=prf_pub"
If you would like to reply directly to Dr Kundaikar, please visit his networker page: https://www.medchemnet.com/users/5115-harish-kundaikar