Preprint Highlight: Probing the rules of cell coordination in live tissues by interpretable machine learning based on graph neural networks

Abstract

Cell–cell interactions are involved in cellular decision making, but there is a lack of systematic methods to quantify and interpret their role. This study presents a computational methodology that predicts epidermal cell fate at the single cell level using information obtained from live cell images. Specifically, the authors used “graph neural networks” to extract information on cells and their neighbors and predicted the cells' future fates with regard to division or delamination. They then identified the features driving the neural network's prediction to infer potential new mechanisms of cell–cell interactions. This unbiased inference of new cell–cell interactions correlated with cell fates is a powerful approach to decipher cell fate decision making in the context of the microenvironment.