Simple Network Mechanism Leads to Quasi-Real Brain Activation Patterns with Drosophila Connectome

Considering the high computational demands of most methods, using network communication models to simulate the brain is a more economical way. However, despite numerous brain network communication models, there is still insufficient evidence that they can effectively replicate the real activation patterns of the brain. Moreover, it remains unclear whether actual network structures are crucial in simulating intelligence. Addressing these issues, we propose a large scale network communication model based on simple rules and design criteria to assess the differences between network models and real situations. We conduct research on the biggest adult Drosophila connectome data set. Experimental results show significant activation in neurons that should respond to stimulus and slight activation in irrelevant ones, which we call quasi-real activation pattern. Besides, when we change the network structure, the quasi-activation patterns disappear. Interestingly, activation regions have shorter network distances to their input neurons, implying that the network structure (not spatial distance) is the core to form brain functionality. In addition, giving the input neurons a unilateral stimulus, we observe a bilateral response, which is consistent with reality. Then we find that both hemispheres have extremely similar statistical indicators. We also develop real-time 3D large spatial network visualization software to observe and document experimental phenomena, filling the software gap. This research reveals network models' power: it can reach the quasi-activation pattern even with simple propagation rules. Besides, it provides evidence that network structure matters in brain activity pattern generation. Future research could fully simulate brain behavior through network models, paving the way for artificial intelligence by developing new propagation rules and optimizing link weights.