On human whole-brain nanoscale connectome: Modeling and data bigness estimation

Knowledge of the human nanoscale connectome is crucial for understanding brain function in health and disease. However, data required to construct a complete nanoscale connectome remain unavailable, and the exact numbers of circuits forming the connectome and neurons within each circuit are unknown. This study (extended conference paper: Nowinski, LNCS15907-Part I:295-309) introduces nanoscale morphologic connectomic wireframe and geometric models, each comprising three sub-models (straight and enhanced with parabolic and cubic branches); provides formulas estimating their data bigness; assesses required storage; proposes model simplification, and benchmarks model size against exascale supercomputers. The connectome size/storage estimation builds upon prior work on the synaptome. To account for the great variability in neuronal and synaptic counts, two estimates for the total number of brain neurons (86/100 billion) and three estimates for synapses-per-neuron (1,000/10,000/30,000) are considered across six connectomic models, yielding 36 storage estimation cases. The straight wireframe model requires from 2.32PB (86 billion neurons/1,000 synapses-per-neuron) to 81PB (100 billion neurons/30,000 synapses-per-neuron). The straight geometric model needs from 2.84PB (86 billion neurons/1,000 synapses-per-neuron) to 99PB (100 billion neurons/30,000 synapses-per-neuron). Model enhancement significantly increases storage from 3.69PB for the parabolic wireframe model (86 billion neurons/1,000 synapses-per-neuron) to 219PB for the cubic geometric model (100 billion neurons/30,000 synapses-per-neuron). Additionally, model simplifications and future research directions are proposed. Modern exascale supercomputers can store any of connectomic models in parallel filesystems, whereas some system memories handle only up to 1,000 synapses-per-neuron cases. This study is the first providing a bigness data estimation to represent the human whole-brain nanoscale connectome.