Increasing k increases the stiffness of the spring, therefore reducing the width of the sampling window at a given distance along the reaction coordinate. Stronger spring = stronger bias towards the window center.

Unfortunately, there are no hard rules here, and the exact values depend on the system and the nature of the interactions in it. The best approach (and one that reviewers like!) is to test several combinations of the parameters to indicate whether any introduce bias or lead to instability. If you can produce a repeatable path with different k or v values, then it shows that the outcome is independent of those settings. Based on that information, you can choose the values that are both practical (in terms of speed of collecting data) and sufficiently accurate.

Can anyone tell me what will happen to the ligand when you increase or decrease the spring constant (k). Especially during the umbrella sampling simulation. Whether lesser K will allow the ligand from one sampling point to the next during the production run or the increased K ?

Increasing k increases the stiffness of the spring, therefore reducing the width of the sampling window at a given distance along the reaction coordinate. Stronger spring = stronger bias towards the window center.

@ Stephane Abel: Thank you so much for the review. It was really useful.

@Justin Lemkkul: Firstly, thank you vey much for the reply. Yes, some papers have done the testing with the range of k and v and finally landed up with the optimum combination where they also say they chose the particular combinaion as the protein rupture was minimal with that combination. Now, my question is, whether it is 50 kcal/mol or 5 kcal /mol , my inhibitor crosses the channel after ceretain time (takes longer in the case of lower k). Here do I determine the rupture that these people are loking at? How do I quantify this rupture in order to choose the right combination? or is there any other parameter that will indicate me about the extent of bias?

How are you defining "rupture?" Do you mean an actual rupture force measurement or are you referring to destabilization of the protein structure? If both 5 and 50 work and don't destabilize anything, I tend to favor the lower force constant to perturb the system in the smallest way possible.

@Justin Lemkul: Dear Sir,

I came across one such article wherein they choose the optimum value of k and v by calculating the rupture forces and this was what I was referring to:

http://pubs.acs.org/doi/full/10.1021/ja302269r

And with reference to your previous answer, how do we calculate the extent of destabilisation ? Is it in terms of system's RMSD?

@Justin Lemkul: As per your suggestion and based on the brief literature search where they hav applied SMD to characterise ligand egress from the binding pocket ( similar to my case) I found the most commonly used k value to be 5 Kcal/mol and v=10 A/ns.

I performed a range of calculations with k=1 kcal/mol. 5 kcal/mol 10kcal/mol and 50 kcal/mol while maintaining v= 10 A/ns In all the cases the drug follows the same path (visually) of egress. But the inhibitor takes less than 3 ns to exit in the latter 3 cases. In that case, if I choose 5 kcal/mol where the force is high enough to pull out within a reasonable time while introducing minimum bias be taken as the final k for my simulations?

Kindly provide your valuable suggestions in this regard.

If the outcome is the same across a variety of conditions, then you have shown it to be independent of the choice of k. You can use whichever setup you like in that case and have justification for when reviewers ask (normally such demonstrations are provided in supplemental information in journal articles).

This is the plot that I am getting by running with k=5 kcal/mol and k=10kcal/mol with v=10 A/ns in both cases. I am unable to interpret this plot and hence land up with a right combination of k,v rationally.

Kindly help me in interpreting this graph of potential mean force vs distance.

Thank you in advance.

The outcomes appear basically indistinguishable. If visualization of the trajectory confirms that they are the same, either setup can be used without issue.

From the provious plot, as a beginner, I am unable to understand your comment that the plots are indistinguishable. Is it based on the height of first energy barrier from the right side ? And what are the other aspects that one needs to focus in this plot?

The other plot of postion vs time is attached herewtih. Will this one provide us any insight in making the choice ?

The question might be trivial but most of the papers where they have taken a range of values of k and v, the plot is usually in the supplementary material where they have not explicitly mentioned the rationale for choosing a particular pair of k and v.

Kindly help me in interpreting this plots.

Thank you so much for your time.

The differences in both plots are minimal; at this point it's splitting hairs. Visualization is the best way to convince yourself of what's going on. If both trajectories show the same general path, there's no appreciable difference. Of course, events will unfold slightly more slowly with k = 5 (as evident from both of the plots), given that the displacement of the spring will have to be larger to overcome the restoring forces within the system. The point is: does the force constant or velocity cause artifacts in the path? If the answer is no, you can choose whatever setup you like.

Thank you so much for your reply. As I had already mentioned earlier, visually, in both the simulations, the ligand follows the same path of egress. The channel that my ligand needs to traverse across is only 15 A while I am using a pulling velocity of 10 A/ns. Is the pulling velocity on the higher side ?

Kindly provide your suggesions and comments.

You velocities are well within what is normally considered standard in such simulations. I would have no concern using either k = 5 or k = 10 in this case. If you're worried about k = 10 being too fast, then use k = 5; your plots above show that you get basically the same answer.

Thank you so much for the all the time you have taken for clarifying my queries and for the valuable explanations on SMD.

Yours sincerely,

Revathi.S

Hi everyone,

What if my force constant is less than 1, there is a problem?

For example, for mine simulation system the value v=0.00005 A/ts and k=0,2 ---          v= 0.000012 and k = 0.2 coincided with the experimental data.

I work with prospecting drugs and I'm balancing my simulation parameters according to experimental data of known binders, and so apply these parameters in simulations with my drugs.

I apologize if I'm using this topic in the wrong way, however, I found that my question is within the theme of this topic