Automated identification of data-consistent spiking neural network models

Variations in cellular and network parameters shape neural dynamics and computation. Mechanistic models, such as spiking neural networks (SNNs), are instrumental for linking recordings of neural population dynamics to their underlying circuit parameters. However, SNNs lack analytical expressions and accessible gradients for parameter inference (i.e., model identification), making it difficult to systematically identify a single configuration that can reproduce experimental data, let alone the full space of data-consistent models and their uncertainty. As a result, SNNs are often reduced a priori to have few free parameters, limiting their potential as mechanistic models of neural circuits and dynamics.

Here, we leverage recent advances in simulation-based inference (SBI) to perform automated model identification for SNNs with high-dimensional parameter spaces using arbitrary features of neural recordings. Simulating 500,000 random configurations of an adaptive exponential integrate-and-fire (AdEx) network with clustered connectivity, we observe complex dynamics with non-trivial dependencies to 28 model parameters representing neuronal, synaptic, and connectivity mechanisms. To enable Bayesian parameter inference, we train a deep neural density estimator on this simulated dataset to approximate the posterior distribution over parameters given observed data.

We first validate our approach by performing inference on simulated network activity with known parameter values. SBI identifies many models (including the ground-truth) that can reproduce both observed and unobserved features of network dynamics, revealing covariance structure and degeneracy between parameters. We then perform inference on real experimental data: applied to a dataset of brain organoid recordings, we identify models consistent with observed network burst features, while elucidating the co-evolution of cellular and network parameters over 40 weeks of development. Together, these results demonstrate how SBI can advance our understanding of the dynamical regimes of flexibly parameterized SNNs, while providing mechanistic explanations and generating hypotheses about hidden circuit properties that underlie changes in brain network dynamics.