An Embodied Computational Model to Explore Environmental-Neural Interactions in Disorders of E/I Balance.

Background: Computational simulations of spontaneous neural dynamics have revealed the importance of local homeostatic excitation/inhibition (E/I) balance mechanisms for the maintenance of flexible neural dynamics in the healthy brain. Further, imbalances in E/I regulation have recently emerged as a leading candidate causal path to autism and common comorbidities like epilepsy. E/I balance is particularly vital during early development, where massive changes to both the structure and function of the brain occur over a relatively short time-period. Maintaining regulation during this period is challenging, and changes in homeostatic set point through alteration of E/I balance may enable waves of new learning. However, current experimental approaches lack a mechanistic understanding of how E/I balance may affect behaviour and brain development. Thus dynamic modelling approaches must be used.

Objectives: Here, we demonstrate a tool for exploring the effect of E/I based homeostatic mechanisms on specific modes of behaviour using an 'embodied' computational model; and consider putative applications of this for the emergence of autism-related traits during development.

Methods: Our approach begins by defining a virtual 'agent' that can move within a 2-dimensional plane, bounded by surrounding walls (Figure 1). ‘Neural’ Dynamics in the model are provided using a range of structurally based computational models, tuned to the specific hypothesis (e.g. [1] (Figure 1A/B). Movement of the agent in the virtual environment is determined by activity within two pre-defined “motor” nodes in the model. Direct manipulation of a group of experimenter-defined task-positive nodes simultaneously enables both “visual” and “somatosensory” inputs to the model - providing an 'open loop' interaction between the dynamics of the 'brain' of the agent and its subsequent manipulation of the environment (Figure 2).

Results: Using this framework we demonstrate the potential of a range of simple manipulations of this model to enable exploration of local and large-scale E/I homeostatic mechanisms during learning and development - particularly those described in [2]. Such an approach raises the possibility of rapidly testing and manipulating hypotheses drawn from rich computational accounts of neural stability dependent on E/I balance. Our demonstration highlights the use of this tool to illustrate that local homeostatic balancing of E/I at the local level enables the emergence of exploratory behavioural dynamics (Figure 3). Moreover, we show that such mechanisms, manipulated during critical points of development lead to long-term alteration in the rich computational dynamics in the brain.

Conclusions: We demonstrate here a role for homeostatic plasticity in local E/I circuits in the emergence of stable exploratory behaviour (trajectories through the environment). Our initial work highlights the possible use of such a tool to explore GABAergic and Glutamatergic models of Autism - providing an in-silico model from which to explore novel treatment approaches. Our longer-term goal is to test predictions from this model within developmental datasets, such as prospective longitudinal cohorts of infants at risk for autism.