You are here: Home News 2019 Disentangling the complex signal …

Disentangling the complex signal flow in layered networks of neocortex

July 22, 2019: Networks in the brain often have a complex structure, comprising multiple layers or otherwise circumscribed subpopulations of nerve cells. Will it ever be possible to untangle the mind-boggling chatter among neurons that experimenters record in such networks with their highly sophisticated methods? Researchers from Stefan Rotter’s research group at the Bernstein Center Freiburg are now proposing a novel theoretical framework to sort out some of the encountered difficulties. Using the example of a computer simulated network model of layered visual cortex, they are able to show, among other things, how feature selectivity and tuning curves of nerve cells in different layers emerge as a result of system-wide processing. Their work has now been published in the journal PLOS Computational Biology.
Disentangling the complex signal flow in layered networks of neocortex

Figure legend see below. Click to view full-size image.


Most nerve cells in all layers of visual cortex respond to spatially extended light bars in a distinctly orientation-dependent way. Cells fire action potentials at a high rate only if the visual stimulus matches their respective “preferred” position and orientation in space. Such feature selectivity, however, is not the same in all neuronal subpopulations of the different layers of visual cortex. There are cell populations that are highly selective, and others that have a more shallow preference. “We wanted to know whether this heterogeneity in firing behavior can be explained by the connectivity of the network within layers and across layers alone”, says Benjamin Merkt, PhD student at the Bernstein Center Freiburg, “or whether just knowing the neuronal circuit is insufficient and other factors must be taken into consideration.”

To generate the connectivity of such networks on a computer, the Freiburg team made use of anatomical information measured and published by other scientists, accounting for several dozen research papers. The resulting mathematical model consisted of almost 80.000 neurons distributed over eight different subpopulations. The model allowed them to simulate the outcome of activity recordings in the visual cortex of rodents that were exposed to visual stimuli. “Our simulations show that differences in orientation selectivity across layers as reported in experiments can be explained by neuronal connectivity alone,” summarizes Benjamin Merkt. “This is a remarkable result.“

Mastering nested feedback loops

The researchers then wanted to investigate in more detail how the structure of the circuit determines its function. “It is often assumed that a signal travels through the different layers of the cortex in a feed-forward fashion until it is eventually transmitted to the next processing stage,” explains Benjamin Merkt. “However, experimentalists have observed phenomena that cannot be understood with this approach at all. Unfortunately, it is not as simple as predicting the activity of one population by the input it receives from another one,” says the young scientist. “Rather, neuronal responses emerge in a dynamic fashion and cannot be easily inferred from static neuroanatomy. In particular, we have strong and nested feedback loops in layered neocortex, resulting in unintuitive behavior.”

The scientists propose a novel system-level approach with which the input-output relations of such strongly recurrent networks can be analyzed. “In our model, we treat the visual cortex as an integrated system and no longer analyze its components separately. We describe what happens in all layers of the network simultaneously, when an input signal is provided. This way, we can even mathematically explain activity patterns that result from multiple nested feedback loops.”

As another result of the new mathematical framework, the researchers also describe a new approach to devise new stimulation experiments. Employing precomputed optogenetic stimuli, for example, specific changes to the activity of a single subpopulation can be achieved, enabling future experiments of unprecedented sophistication.

The current version of the model is based on several simplifications. The neurons, for example, all have the same properties – a feature, which does not reflect well the biological reality. “We are nevertheless convinced that we have devised a useful and robust new analysis method that is well suited for the analysis of complex neuronal systems,” Benjamin Merkt sums up. “It can be applied to a wide range of neuronal networks, and applications to other systems are already under way.”

Figure Legend
Layered network of the primary visual cortex of rodents. Massive inter-layer connectivity leads to a specific spread of neuronal preferences for stimulus orientation across layers. New modeling approaches are used to analyze the signal flow in such networks, revealing a very good match between computer simulations and spike trains recorded in animals.

Original Publication
Merkt B, Schüßler F, Rotter S. Propagation of orientation selectivity in a spiking network model of layered primary visual cortex. PLOS Computational Biology, 2019.

https://doi.org/10.1371/journal.pcbi.1007080

Read also (Portrait)
A constant thirst for new knowledge

Contact
Benjamin Merkt
via
Prof. Dr. Stefan Rotter
Professor of Computational Neuroscience
University of Freiburg
Bernstein Center Freiburg & Faculty of Biology
Hansastraße 9a
79104 Freiburg, Germany
Tel.: +49 (0)761 203 9316
E-mail: stefan.rotter@bio.uni-freiburg.de

 

 

Latest News

Back to overview

All

20242023 | 202220212020 | 2019 | 2018 | 2017 | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009