hide
Free keywords:
-
Abstract:
Network bursting is a common type of collective dynamics that robustly emerges in in vitro networks of dissociated primary neurons, cultures of IPSC-derived neurons, and brain organoids [Sukenik et al. 2021]. The mechanisms of network bursting are not yet fully understood. Here we analyze a simple macroscopic rate model with firing rate adaptation that matches the bursting dynamics statistics in vitro. We show that in vitro-like bursting can arise from one of three possible mechanisms: noise-driven bistability, limit cycles, or in the excitable state. Using the model to predict changes in the behavior under pharmacological manipulations allows testing the generality of the model and identifying the most probable dynamical state.
We focus on a simplified macroscopic model that includes a recurrent unit with a sigmoid nonlinearity and slow activity-dependent adaptation mechanism. We approximate the posterior distribution of the model parameters given the statistics of inter-bursting intervals and average burst duration of hippocampal cultures by fitting a conditional density estimator based on neural spline flow [Durkan et al. 2019]. The resulting posterior distribution shows that the most probable parameters fall onto the excitable state. We also found that excitability and adaptation can compensate for each other, allowing the model to exhibit indistinguishable bursting dynamics for a wide range of parameters.
One of the model's predictions is that inter-burst intervals become smaller as excitability increases, whereas burst durations increase.
