Brain-wide dendrites in a near-optimal performance of dynamic range and information transmission

Abstract Dendrites receive and process signals from other neurons, and transfer the information to soma. The range of signal intensities that can be processed is referred as dynamic range. We model dendrites in a neuron as multiple excitable binary trees connected to a common soma and each node in dendrites can be excited by external stimulus or by receiving signals transmitted from adjacent excited nodes. We find that even with equal forward and backward transmission probability through dendrites, double sigmoid behavior can appear for neurons with a large number of somatic branches. We show that the dynamic range of dendrites increases with the interaction probability and the number of somatic branches, and decreases with the asymmetry of dendrites. Moreover, we show that neurons with more somatic branches perform with low energy costs and more efficiency for information transmission. However, further increasing the number of somatic branches (e.g. beyond 10 somatic branches), has limited ability to improve the transmission efficiency; with brain-wide neuron digital reconstructions of the pyramidal cells, 90% of neurons have no more than 10 dendrites. These suggest that actual dendrite morphology is near optimal with an intermediate number of somatic branches, in terms of dynamic range and information transmission.