Interactions between long- and short-term synaptic plasticity transform temporal neural representations into spatial patterns

An international research team led by Professor Robert Gütig from the Berlin Institute of Health at Charité (BIH) and Charité – Universitätsmedizin Berlin has discovered that the flexible coordination of short- and long-term plasticity enables nerve cells to process temporal sequences of action potentials as spatial patterns of activity. This mechanism increases the capacity and reliability of neuronal circuits and may explain how the brain produces complex learning and memory functions. The study has now been published in PNAS.

Information processing in the brain relies on the transmission of spikes through chemical synapses whose efficacy often depends on their recent firing history. While the effects of short-term plasticity on neural processing have been extensively studied, it remains unclear how interactions between short- and long-term plasticity influence the learning capabilities of neural networks. Recent findings demonstrate that long-term modifications of short-term plasticity at individual synapses enable neurons to learn to process temporal spike sequences as if they were spatial patterns. This mechanism enhances the capacity and robustness of neural circuits, albeit at the cost of increased spiking activity. To double the storage capacity of a neural network one would not need to double the number of connections as in conventional neural networks, but rather, it would be sufficient to double the number of action potentials.

These findings are based on phenomenological models of short-term plasticity and fit recent electrophysiological measurements of brain activity. The theory predicts that the learning rule at a given synapse depends on the degree and type of short-term plasticity induced by long-term plasticity induction protocols.

New Insights into the Dynamics of Neural Learning Processes

“These results provide new insights into how the brain can process information flexibly and efficiently,” says Robert Gütig, W3 Professor in Mathematical Modeling of Neural Learning at Charité and the BIH and deputy director of the BIH Charité Digital Clinician Scientist Program. “By integrating short- and long-term plasticity, neural networks can transform temporal sequences of spikes into spatial patterns — a crucial step for complex learning and memory formation.”

Further research may exploit these mechanisms for advances in neuroscience and artificial intelligence, for instance to boost the performance of neuromorphic network architectures, that can operate at particularly low levels of power consumption.

The figure shows an example of the gradual transition from an inhibitory synaptic response (blue, cool colors) for short time intervals between incoming action potentials to an excitatory signal transmission (red, warm colors) for longer time intervals. Copyright: Robert Gütig

Original publication:
Q. Yu,M. Tsodyks,H. Sompolinsky,D. Schmitz, & R. Gütig, Interactions between long- and short-term synaptic plasticity transform temporal neural representations into spatial, Proc. Natl. Acad. Sci. U.S.A. 122 (47) e2426290122, https://doi.org/10.1073/pnas.2426290122 (2025)

Source:
Press release of the Berlin Institute of Health (BIH)

Contact:
Prof. Dr. Robert Gütig
Charité – Universitätsmedizin Berlin
E-mail: robert.guetig[at]bih-charite.de