cluster jointly with morphology and connectivity

[bdpedigo]

• can use MASE to jointly embed a graph and pairwise NBLAST matrix
• can compute discriminability to measure cell type differentiation if there is labelled data
• labeled data is rarely perfect and usually biased towards the morphology

Example of this approach on the mushroom body-ish subset of data: https://docs.neurodata.io/notebooks/pedigo/maggot/graspologic/2020/12/15/joint-connectivity-morphology.html

• currently working on this on the full dataset. metrics become less clear in that case.

[bdpedigo]

• can also run this on the early olfactory system datasets?
  • I think only part of them will have connectivity also