Bastian Eppler from Matthias Kaschube's FIAS research group defended his thesis on 19 July and celebrated with the team afterwards. In his doctorate, he studied the mechanisms of how the brain stores information stably over long periods of time and at the same time integrates important new information quickly.
"We investigated how stable activity patterns are at the level of neuronal populations without external influences," Eppler explains, "and how they change during learning." In addition, he wanted to find out how changes in the network connectivity influence the network´s activity. To do this, he analysed the theoretical model of a neuronal network.
Using data of neuronal activity in mice, Eppler showed that neuronal activity changes dynamically all the time, even in the absence of any learning paradigm. Classic conditioning biases these ongoing dynamic towards selective generalisation, which is also found in the animals' behaviour. This is similar to PTSD (Post Traumatic Stress Disorder, see News / FIAS).
He also used a network model to investigate how changes in network connectivity affect the network activity. Apparently, constant changes in connectivity lead to phases of stable activity, but these are interrupted by abrupt transitions to another network activity. These abrupt transitions coincide with certain qualitative changes in the fixed point topology of the network.
Neuronal population activity is not as stable as previously assumed. It is undergoing constant remodeling. These ongoing changes are utilized and biased during learning. In a model these changes were abrupt and were mitigated by changes in the network's fixed point structure.
Eppler will continue his studies at FIAS as a postdoc for the time being - while looking for another postdoc position.
Thesis 2022: Bastian Eppler, Ongoing neuronal population activity dynamics in the neocortex - representational drift in experiment and model.
Dominik F. Aschauer, Jens-Bastian Eppler, Luke Ewig, Anna R. Chambers, Christoph Pokorny, Matthias Kaschube, Simon Rumpel, Learning-induced biases in the ongoing dynamics of sensory representations predict stimulus generalization, Cell Reports, Volume 38, Issue 6, 2022, 110340, https://doi.org/10.1016/j.celrep.2022.110340