Identification via Numerical Computation of Transcriptional Determinants of a Cell Phenotype Decision Making

Complex cellular processes, such as phenotype decision making, are exceedingly difficult to analyze experimentally, due to the multiple-layer regulation of gene expression and the intercellular variability referred to as biological noise. Moreover, the heterogeneous experimental approaches used to investigate distinct macromolecular species, and their intrinsic differential time-scale dynamics, add further intricacy to the general picture of the physiological phenomenon. In this respect, a computational representation of the cellular functions of interest can be used to extract relevant information, being able to highlight meaningful active markers within the plethora of actors forming an active molecular network. The multiscale power of such an approach can also provide meaningful descriptions for both population and single-cell level events. To validate this paradigm a Boolean and a Markov model were combined to identify, in an objective and user-independent manner, a signature of genes recapitulating epithelial to mesenchymal transition in-vitro. The predictions of the model are in agreement with experimental data and revealed how the expression of specific molecular markers is related to distinct cell behaviors. The presented method strengthens the evidence of a role for computational representation of active molecular networks to gain insight into cellular physiology and as a general approach for integrating in-silico/in-vitro study of complex cell population dynamics to identify their most relevant drivers.