Deep learning-based synaptic event detection

O'Neill, P. S.; Baccino-Calace, M.; Rupprecht, P.; Friedrich, R. W.; Mueller, M.; Delvendahl, I.

Quantitative information about synaptic transmission is key to our understanding of neural function. Spontaneous synaptic events carry important information about synaptic efficacy and plasticity. However, due to their stochastic nature and low signal-to-noise ratio, reliable and consistent detection of these events in neurophysiological data remains highly challenging. Here, we present miniML, a novel method for the accurate detection of spontaneous synaptic events based on deep learning. Using simulated ground-truth data, we demonstrate that miniML outperforms commonly used methods in terms of precision and recall over different signal-to-noise conditions. The event detection method generalizes easily to diverse synaptic preparations and different types of data. miniML provides a powerful and easy-to-use deep learning framework for automated, standardized and precise analysis of synaptic events in any cell, thus opening new avenues for in-depth investigations into the synaptic basis of neural function and dysfunction.