less random 3D point selection : Calculate Points within a Cylinder

[kellnett]

Currently, the way we randomize 3D points may be too random since dendritic spines do have biological clustering mechanisms. I just want to look at whether the treatment alters the degree of clustering.

One way we could possibly select less random points is by having a criteria for coordinate inclusions. For example, use the maximum spine length of that section as the radius of a cylinder and only accept points within that volume.

[kellnett]

The next solution could be using the current spine attachment points (which we have as 3D coordinates) as the origin of a sphere with a radius equal to the maximum spine length and generate a random 3D coordinate within that spherical volume for each attachment point along the dendrite.

This could possibly help control for any biological process that determines where the spines begin to form (since my data do show some degree of attachment point clustering)

[kellnett]

This is really dreaming big, but if we could somehow use control dendritic spine samples along with machine learning to find an algorithm for natural spine clustering. Then we can compare our treatment group to the expected clustering to see if our observed sample is more or less clustered.

[jdkent]

I have a link to my current thought process about this option, but we can continue the conversation through this issue.

[kellnett]

I think I understand your point. Basically, because there is more possible space for a random point to be (even within a cylinder) that the likelihood of any of those random points being clustered is lower than what is biologically seen?

If that's the case then comparing to control data does seem like the best bet.