less random 3D point selection : Machine Learning

[jdkent]

@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]

This is another interesting option, but for now I believe this would be conceptually similar to doing a simple contrast between the controls and treatment groups to see how much they differ. Currently, to me, generating a model to represent what controls "look like" and then contrast with the treatment group is a round about way of testing how controls look different from treatment.

Happy to be persuaded otherwise