Membrane curvature for a dynamic system

[rajiv-venkatesh]

The MDAnalysis membrane-curvature tool uses fixed bins when generating curvature heatmaps. The size of these bins is based on the initial size of the system. If my system grows over the trajectory, would it be best to use only the part of the trajectory with stable size, or is there a way to dynamically resize the heatmap bins?

My system is initially 300 x 300, and I put a 30x30 grid for my curvature heatmap. However, in the first 20 ns of my production run, it grows to 350 x 350 A. The grid squares are actually capturing different patches of membrane over time as a result. Would it instead be best to begin analysis when the system's size is stable?

[p-j-smith]

Hi @rajiv-venkatesh

Would it instead be best to begin analysis when the system's size is stable?

Yep exactly - this is a good way of checking that your system has equilibrated. It's best if you discard this equilibration period for any analysis you do, not just the membrane curvature.

[rajiv-venkatesh]

Hi @p-j-smith, thank you for your response!

I have a 150 ns trajectory of a protein embedded in lipid. When I try to run the mean curvature calculation on my system, I observe white patches on the curvature heatmap that are due to some bins having no value (na). I have attached a picture of the heatmap. My system's lipid size is 320 x 320 A, and I am using a 30 x 30 bin grid. I do not set any x_size and y_size in the MembraneCurvature() function at the start of the script. Is there a reason for these white patches?

[p-j-smith]

The empty patches are where your protein is inserted. There's an issue in the membrane-curvature repository around whether these patches should be interpolated or kept empty. I would rather keep the patches empty, but it's your decision.

Although it looks like the empty patches are not where you expected (assuming the green dots represent your protein?). Did you align your trajectory before running the analysis?

[rajiv-venkatesh]

Thanks for the quick reply!

I do not think I aligned my trajectory as described at the link, so I can try that. It is odd that the empty patches are not where the protein is located. I find that when I make my bin sizes smaller (20x20 instead of 30x30) the patches disappear—I would very much prefer the 30x30 bins, though. Still, I don't think that there are frames where there are no lipids in these 10 A x 10 A boxes. Could there be another cause?