Multi-output computation by single neuron biophysics in a visual system

Abstract

As long anticipated (Sandberg and Bostrom 2008; Seung 2012; Szigeti et al. 2014), connectomics is providing a new foundation for brain simulation by replacing theoretical assumptions about network connectivity with solid empirical facts. Connectomics also yields detailed information about neuronal morphology, which is useful for simulating the biophysics of single neurons (Yang et al. 2016; Meier and Borst 2019; T. X. Liu et al. 2022). Here I introduce a formalism for simulating a brain as a network of synapses interacting via an effective resistance matrix. By computing this matrix for fly visual interneurons, I find evidence that some neurons may be true multi-output devices, neither well approximated as "point neurons" (Lappalainen et al. 2024; Shiu et al. 2024), nor as collections of functionally independent compartments (Meier and Borst 2019). Within a linear approximation, such a neuron is instead equivalent to a hierarchy of virtual neurons that spatially pool over multiple length scales. The computational powers of multi-output neurons may support highly sophisticated normalizations in the fly visual system (Seung 2024a).