Bayesian Optimisation of Large-Scale Biophysical Networks

Abstract

The relationship between structure and function in the human brain is well established, but not yet well characterised. Large-scale biophysical models allow us to investigate this relationship, by leveraging structural information (e.g. derived from diffusion tractography) in order to couple dynamical models of local neuronal activity into networks of interacting regions distributed across the cortex. In practice however, these models are difficult to parametrise, and their simulation is often delicate and computationally expensive. This undermines the experimental aspect of scientific modelling, and stands in the way of comparing different parametrisations, network architectures, or models in general, with confidence. Here, we advocate the use of Bayesian optimisation for assessing the capabilities of biophysical network models, given a set of desired properties (e.g. band-specific functional connectivity); and in turn the use of this assessment as a principled basis for incremental modelling and model comparison. We adapt an optimisation method designed to cope with costly, high-dimensional, non-convex problems, and demonstrate its use and effectiveness. Using five parameters controlling key aspects of our model, we find that this method is able to converge to regions of high functional similarity with real MEG data, with very few samples given the number of parameters, without getting stuck in local extrema, and while building and exploiting a map of uncertainty defined smoothly across the parameter space. We compare the results obtained using different methods of structural connectivity estimation from diffusion tractography, and find that one method leads to better simulations.

A Connectome-Based, Corticothalamic Model of State- and Stimulation-Dependent Modulation of Rhythmic Neural Activity and Connectivity

“…Notably, this network-based approach allowed us to harmonize the analysis of functional connectivity in simulated and empirical MEG data. In this we followed the approach of Abeysuriya et al (2018) and Hadida et al (2018) in our use of the bandpass-filtered amplitude envelope correlations (Brookes et al, 2011;Hunt et al, 2016), and that line of work is perhaps the closest of recent modelling studies to the present one. Abeysuriya et al (2018) studied the role of inhibitory synaptic plasticity in a connectome-based network of Wilson-Cowan equations.…”

A connectome-based, corticothalamic model of state- and stimulation-dependent modulation of rhythmic neural activity and connectivity

“…Notably, this network-based approach allowed us to harmonize the analysis of functional connectivity in simulated and empirical MEG data. In this we followed the approach of [36] and [87] in our use of the bandpass-filtered amplitude envelope correlations[78, 37], and that line of work is perhaps the closest of recent modelling studies to the present one. In [36], Abey-suriya and colleages studied the role of inhibitory synaptic plasticity in a connectome-based network of Wilson-Cowan equations.…”

Model Based Inference of Large Scale Brain Networks with Approximate Bayesian Computation

“…Ritter et al, 2013). Recent work developing optimization of large scale neural models has shown that by improving parameter estimation algorithms, it is possible to fit high-dimensional, complex models of neural dynamics without appealing to steady-state approximations (Hadida et al, 2018). By relaxing the restrictions on model dynamics, it is thus possible to fit a model using a combination of a much wider range of data features beyond that contained in just the cross spectral density.…”

“…In future work, a test of construct validity should be conducted to examine the consistency of model estimates made between the ABC based method and existing approaches. Current methods adopting algorithms such as variational Laplace (Friston et al, 2007); deterministic sampling approaches (Hadida et al, 2018); and Generative Adversarial Networks (Arakaki et al, 2019) all provide potential routes to estimation of network mechanisms of spontaneous dynamics yet each is likely suited to different pairings of data and models. A quantitative assessment of their applicability in different scenarios would provide concrete answers to the inevitable question of which scheme is best is for a particular type of problem.…”