How does network structure shape brain activity? : New article in PLoS Computational Biology relates correlations in the activity of nerve cells to the detailed structure of the underlying network
Click to enlarge. See image caption below.
Although origin and function of correlations are not known in detail, they are believed to play a prominent role for information processing and learning. It was known that correlations can be caused, for instance, by direct interaction between cells due to a synaptic coupling between them. More importantly even, correlations can arise due to shared input that two cells receive from a third cell, or a whole population of cells that represent a common source. It is conceivable, though, that many other types of indirect interactions among cells contribute to the correlations measured in experiments as well. Therefore, in order to make sense of the multi-neuron data that are recorded from the brain with modern methods, and to relate them to the neuronal tissue that originated these data, elaborate analyses of dynamic neuronal networks are necessary.
Scientists at the Bernstein Center Freiburg have now made a significant contribution to this endeavor and present their work in the journal “PLoS Computational Biology”. Volker Pernice and his colleagues have devised a method to relate pairwise correlations in the activity of spiking nerve cells to the detailed structure of the underlying network. The researchers achieved this by systematically accounting for all the various routes of synaptic connectivity encountered in a network that end up on one particular neuron. For many networks of the brain, the analysis is somewhat complicated by the fact that the output of a cell, after having passed through several intermediary cells, can act as an input to the same cell. Such network structures are known as “recurrent”. The researchers outlined the conditions under which the combined effect of all indirect interactions – via pathways that may include several synaptic transmission steps – make significant contributions to shaping correlated activity.
This work also demonstrates that the microstructure of brain networks, the uncovering of which is subject of the newly emerging field of “connectomics", exerts a strong and specific influence on the resulting activity dynamics.
As the scientists point out in their paper, considerable efforts are currently dedicated to the reconstruction of detailed wiring diagrams of brains on multiple scales. In this context, Pernice and his colleagues are convinced: models that can faithfully reflect the influence of such detailed connectivity will play an important role in future brain research.
Neurons in a network are, apart from direct connections, linked via a large number of indirect connections which contribute to spike train correlations.
2011 How Structure Determines Correlations in Neuronal Networks. PLoS Comput Biol 7(5): e1002059. doi:10.1371/journal.pcbi.1002059