Why have connectome projects failed so consistently in generating any useful models to predict behavior?
Because connectome models are not maps of nervous systems, they are maps of a tiny sliver of neuron specific function. Nearly all connectome models at this point carry the assumptive burden that a) neurons are the primary units of calculation in a nervous system, b) that disparate connections have more weight than local connections, and c) that the extremely limited type of data we collect is representative of the full range of signal activity occurring in the cell.
Connectome projects desperately need to update their fundamental assumptions and at the very least start looking at the contribution of glial populations and “micro-clusters”.
It would be super fascinating to see something like a “metabolome” map pop up in the same vein as connectome projects, where with equivalent resolution we could track the flows of energy through a nervous system, even better if it was across multiple modalities like IR/thermal, Glu, ATP, etc.
Energy is ultimately the difference between something which is inert and something which is active, regardless of it’s construction (or morphology). Following the energy seems like a pretty promising path to improve our ability to understand the mechanics of nervous systems.