A dynamic neural network model of the zebrafish posterior lateral line sensorimotor pathway

Neural network architecture is an important area of study in Neuroscience, as the possible dynamics of a network are highly dependent on its structure. I apply artificial neural networks, tools from the field of Computer Science, to study the neural anatomy of a larval zebrafish Posterior Lateral Line sensorimotor pathway. The goal of this study is to understand the structural constraints a simulated escape response and a simple swimming motion place on network architecture. An artificial neural network, together with realistic stimulus/ response pairs, is used to simulate a simplified zebrafish brain together with escape and swimming behaviors. To characterize the structure of connections in a functioning network I examined the distribution of participating neurons in the model brain. I further examined the effects of increasing network size and the number of stimulus/response pairs on model convergence time and the number of brain neurons participating in activation. I determined theoretically that 4 brain neurons were always sufficient in generating correct escape and swimming responses. This corresponds closely with the Mauthner series in zebrafish thought to be responsible for initiating the escape response. The stimulus/response pairs impose no other significant structural constraints, allowing me to conclude that the highly conserved structure of the zebrafish PLL must result from other processes.