Calculating Average Structure

[Jonathan Saboury]

I want to find the average structure of triphenylmethanol in water from a simulation I have done. Am I right to assume that the functions "ThreeParticleAverageSite()" and "TwoParticleAverageSite" are the ones I am looking for? What about atoms that are connected to four atoms?

I've looked at 21.4 Virtual Sites in the application guide and it seems it is what I need, but I have no idea how to code this to get a single slide pdb with the average structure of the target molecule. Do I need to code it in python and run it? If so what is the general template?

Thanks, your help has been invaluable to the research I am doing and is very much appreciated!

[Peter Eastman]

Hi Jonathan,

Those classes have nothing to do with computing average structures. They're for implementing virtual sites: interaction sites in a molecule (like the two extra charges in TIP5P water) that have no mass, and whose position is calculated based on the positions of other atoms.

I gather what you want to do is take a series of structures recorded during a simulation and compute a single "average" structure from them? That's actually not a trivial thing to do. You could, of course, just calculate the average position of each atom independently, but that probably won't give you anything that looks like a valid structure. Even defining what is meant by an "average structure" can be difficult.

Could you give more details on exactly what you want to accomplish?

Peter

[Jonathan Saboury]

Hi Peter,

Ah, ok.

What I want is exactly what you described. Average structure to me is taking the sample frames and finding the average structure (having trouble defining it further). I guess I could give the comparison of "law of large numbers". The more samples I have the more accurate the structure will be.

The reason I am doing this is so that I can more accurately calculate bond angles and see if there are pi-pi interactions or hydrogen bonding in my product of interest.

I have limited math experience, but just thinking about it calculating the average position of each atom would be able to do what I want, crude, but it would work...I think. I could just create a program for this, or if you have a predefined function that would save me time

Thanks!

-Jonathan

[Jonathan Saboury]

I see what you mean, that obviously didn't work, I didn't think about rotation, silly me.

I was looking into other programs solutions and they all seem the same, they just average the coordinates.

This surprises me because of all this linear algebra/complex math used in the simulations it at least seems to me very little effort was put into averaging the structure (perhaps it is not important in biochem?).

How would you normally prove there was pi-pi or hydrogen bonding happening?

Thanks, Jonathan