Method for identifying cellular states and determining their quantity during epithelial-mesenchymal transitions

Abstract

In complex biological dynamic systems, system states are often associated with stable behavioral modes (e.g., attractors) or characteristic process phases. Biological systems are high-dimensional objects; their states are characterized by a multitude of parameters. This fact makes the task of detecting such states challenging, and considering low-dimensional system portraits often fails to detect system states. Articles devoted to the detection and visualization of such states typically employ two- or three-dimensional images. Detecting such states from multidimensional time series data is a key task in mathematical biology and biophysics. We proposed using the SOM artificial neural network for state detection and visualization. In this work, we analyzed the process of epithelial-mesenchymal transition, which determines cellular plasticity and plays a crucial role in cell and tissue biology, as well as in such important pathologies as cancer. We demonstrated that the EMT dynamics contain six states in a 16-dimensional space. The proposed method allows one to identify and interpret the states of multidimensional biological systems, which can be used in the search for methods of controlling these states.