Figure 13.

The qualitative performance of the Deep-Channel with (previously) unseen data. (a–c) represent examples of Deep-Channel classification performance with ion channels having low activity (data from model M1, Figure 13a,b): (a) raw semi-simulated data of the channel event (in black). (b) ground truth idealisation/annotation labels (in blue) from the raw data in (a). (c) Deep-Channel predictions (in red) for the raw data (a). (d–f) representative examples of Deep-Channel classification performance with five ion channel openings simultaneously (datafrom model M2, Figure 13c,d). (d) semi-simulated raw channel events data (black). (e) ground truth idealisation/annotation labels (in blue) using raw data in (d). (f) the Deep-Channel label predictions (in red) for raw data (d).