A Machine Learning Approach for Predicting Electrophysiological Responses in Genetically Modified HEK Cells

Electroporation involves exposing biological cells and tissues to pulsed electric fields (PEFs). While these electroporation-based techniques are gaining prominence across various application domains, several studies demonstrated that the electric pulses can affect cell excitability, leading to undesirable effects. However, comprehending the interplay between electroporation and the electrophysiological response of excitable cells poses a significant challenge. In this study, genetically engineered human embryonic kidney (HEK) cells were employed as a simple cell model to investigate the effects of electroporation on excitable cells. These cells were exposed to nine 100 mu s pulses of increasing electric field amplitude, cell-wise manually segmented. The resulting binary masks of each cell underwent in a single pulse are organised into a dataset. The objective of this research is to develop an algorithm capable to automatically predict which kind of fluorescence response the single cell produces starting from the binary mask and related experimental informations. Three distinct approaches were implemented namely, a deep neural network, traditional machine learning models, and an hybrid combination of the two. The traditional machine learning model, using manually extracted features, exhibited superior performance among all models.