A deep learning approach to real-time Markov modeling of ion channel gating

The patch-clamp technique allows us to eavesdrop the gating behavior of individual ion channels with unprecedented temporal resolution. The signals arise from conformational changes of the channel protein as it makes rapid transitions between conducting and non-conducting states. However, unambiguous analysis of single-channel datasets is challenging given the inadvertently low signal-to-noise ratio as well as signal distortions caused by low-pass filtering. Ion channel kinetics are typically described using hidden Markov models (HMM), which allow conclusions on the inner workings of the protein. In this study, we present a Deep Learning approach for extracting models from single-channel recordings. Two-dimensional dwell-time histograms are computed from the idealized time series and are subsequently analyzed by two neural networks, that have been trained on simulated datasets, to determine the topology and the transition rates of the HMM. We show that this method is robust regarding noise and gating events beyond the corner frequency of the low-pass filter. In addition, we propose a method to evaluate the goodness of a predicted model by re-simulating the prediction. Finally, we tested the algorithm with data recorded on a patch-clamp setup. In principle, it meets the requirements for model extraction during an ongoing recording session in real-time. The patch-clamp technique enables probing of the gating behavior of individual ion channel proteins, but unambiguous analysis of single-channel datasets is challenging. Here, the authors present a deep learning approach for extracting Markov models from single-channel recordings.