Details of connectivity play important role for network activity

Each neuron in the brain receives input signals from thousands of other neurons, and in turn contributes to the activity of just as many neighbours. In this way, the brain integrates and redistributes information throughout the network of its nerve cells.

The details of these connection patterns among neurons in biological networks are largely unknown, and there is no easy way of directly observing them. It seems very plausible that the network structure influences the statistics of the pulse patterns produced by all the neurons in the network in a collaborative effort. Still, it is difficult to tell how exactly connections between individual nerve cells or groups of them affect the observable activity. Stepping away from the real brain and addressing this question within computer simulations of artificial networks helps to address – and partly answer – this question.

In the simplest conceivable model of a brain network its elements connect randomly and independently of each other. Such networks have been extensively studied in various fields of science, including mathematical graph theory and neurobiology, and their properties are well known. Despite the simplicity of these networks, the activity patterns produced by them share many features with the ones observed in biological networks.

To find out how non-random features in the network structure affect its activity, researchers from the Bernstein Center Freiburg systematically investigated this question with the help of large-scale computer simulations and statistical analyses. Because so little is known about the properties of the brain’s networks, and to make as few assumptions as possible about the way neurons are connected, the scientists first generated a large variety of networks with highly variable circuit layouts. The simulated neural activity of these networks differed strongly, for example depending on the statistical properties of outgoing and incoming connections of single neurons. Other aspects, like the clustering of small cell assemblies, were surprisingly found to have only weak effects on the population activity.

The study, published in Frontiers in Computational Neuroscience, highlights potentially important properties of network models and their mutual dependencies. The researchers hope that the new insights gathered from this theoretical study will also guide the future search for corresponding properties in networks of biological nerve cells.

Figure Legend

Different network structures (columns) produce activity patterns with differences in the spike train statistics of the neuron population (top), which are reflected in the dynamics of individual neurons (middle) as well as their combined activity (bottom).
 

Original article:

Pernice V, Deger M, Cardanobile S and Rotter S (2013) The relevance of network micro-structure for neural dynamics. Front. Comput. Neurosci. 7:72. doi: 10.3389/fncom.2013.00072