A neural network reconstructed an electrical model of a cardiac cell from a single recording
A neural network reconstructed an electrical model of a cardiac cell from a single recording
On July 23, eLife published a study of cardiac cells grown from reprogrammed stem cells. The authors recorded how two living cells responded to a specially designed sequence of voltage commands and used each recording to build a computational model.
A cardiomyocyte contracts in response to an electrical impulse. Ion channels in the cell membrane generate this impulse. Some allow sodium and calcium to enter the cell, while others allow potassium to leave and return the voltage to its baseline level. A conventional recording captures the combined effect of these channels. To explain the shape of the impulse, a researcher needs the individual properties of each current.
In the eLife paper, Pei-Chi Yang's team first created 1,1 million synthetic cardiomyocyte models. They varied 52 parameters across six ionic currents and calculated how each cell would respond to different voltage commands. The authors used this population to select a voltage clamp protocol, in which an instrument holds the membrane at a specified voltage and measures the current elicited by each command.
The neural network learned to infer the parameters that had generated the synthetic recordings. The researchers then gave it a recording from a living cell and obtained a set of parameters for a model of that cell. From a single electrical recording, the authors determined the properties of six ionic currents in an individual cell. In their calculations, the resulting model reproduced the cell's measured electrical impulse.
The authors tested this approach on two cardiomyocytes from the same iPSC line. These cells are produced by reprogramming an adult cell into a stem cell and then directing it toward a cardiac muscle lineage. Using the reconstructed models, the team also calculated the movement of calcium within the cell, which triggers contraction.
The idea of modeling variation among cardiac cells predates this study. In a 2013 PNAS paper, researchers fitted a population of models to experiments on rabbit Purkinje fibers and compared the predictions with responses to four concentrations of dofetilide. The new study automates the inverse process, moving from a single recording of a human iPSC cell to the parameters of its model.
The authors establish a cycle in which a cellular measurement defines a model, and the model specifies a question for the next experiment. The parameters of the six currents turn the shape of the electrical impulse into a testable hypothesis about the mechanism underlying the cell's behavior.