GENERATION OF COHERENT ACTIVITY PATTERNS IN FORCE-TRAINED BIOLOGICAL NEURAL NETWORKS WITH MODULAR ORGANIZATION

In vitro biological neural networks (BNNs) provide a well-defined model system to investigate how living cells interact with their environment by constructing and experimentally manipulating the network to shape high-dimensional dynamics that could be used to generate coherent temporal outputs, such as those required for motor control. Previously, we demonstrated that modular BNNs fabricated using primary rat cortical neurons and microfluidic devices function as reservoir layers in a reservoir computing framework, enabling the generation of periodic and chaotic waveforms. However, the mechanisms by which modular connectivity embedded in BNNs improve signal generation performance remain unclear. In this study, we performed waveform generation experiments on modular BNNs of primary cultured rat cortical neurons and then perturbed (i) interwell connectivity by structurally disconnecting the microchannels and (ii) excitatory synaptic transmission by applying CNQX. In both conditions, waveform generation performance decreased; in particular, under CNQX, the mean squared error between target and output signals increased by 232% relative to control. These results emphasize that modular organization realized by both excitatory synaptic transmission within BNNs and interwell connectivity realized by the microchannels are critical for proper function as the reservoir layer capable of generating coherent temporal pattern. This work offers a biologically inspired platform for understanding the physical basis of cortical computation and for advancing energy-efficient neuromorphic computation. The study is funded by MEXT Grant-in-Aid for Transformative Research Areas (A) “Multicellular Neurobiocomputing,” JSPS KAKENHI, RIEC Cooperative Research Projects at Tohoku University, and the WISE Program for AI Electronics by Tohoku University.