High performance, large-scale multi-compartment Hodgkin-Huxley simulation of Drosophila's whole-brain neural circuit model

A major challenge in neurosciences is the elucidation of neural mechanisms in brains that are crucial for the processing of sensory information and the generation of adaptive behavior. In conjunction with the ever-growing body of experimental data, computational simulations have become crucial in integrating information and testing hypotheses, requiring fast large-scale simulators. We constructed a whole-brain neural circuit model of the fly Drosophila with biophysically detailed multi-compartment Hodgkin-Huxley models based on the morphologies of individual neurons published in open databases. Performance tuning of the simulator enabled near real-time simulation of the resting state of the Drosophila whole-brain model in the large-scale computational environment of the supercomputer Fugaku, for which we achieved in excess of 630 TFLOPS using 480k cores. In our whole-brain model, neural circuit dynamics related to a standard insect learning paradigm, the association of taste rewards with odors could be simulated.