47 — Inference of topology and the nature of synapses, and the flow of information in neuronal networks

This work focuses on the characterization of synapse topology — namely, direction — in a large biological neuronal network. Most synapses are chemical, and therefore act as directed edges in the connectivity graph of the brain. Rare synapses are electrical (un-directed). This work inspects Hindmarsh-Rose neurons — a simulated model that exhibits the burst-spike activity of a membrane in a biological neuron.

Borges et al aim to develop tests to determine the directionality of synapses in a neural network so that one can determine graph topology using only spike-timing information. This is important because anatomical studies are expensive and difficult, and result in huge amounts of imagery data; whereas Ca-imaging or ephys studies result in very small time-series data.

They approach this problem armed with the causal mutual information model (CaMI), which (roughly) states that if information flows from node $A$ to node $B$, then time-series data collected on $B$ can be used to inform a viewer about qualities of $A$. In particular, one can use this information to map long time-series data about $B$ to much shorter time-series data about $A$.

Using CaMI alongside existing other methods, Borges et al demonstrated that they can successfully map individual neurons by synapse type and connectivity direction, even in the presence of Gaussian noise. They also prove that this method has the ability to calculate “population”-level connectivity, if individual-neuron-level precision is not available (e.g. EEG or lower spatial-resolution imagery).