Cells are governed by complex and multi-layered gene regulatory networks, which orchestrate the development of specialized cell types from pluripotent stem cells. Computational models of gene regulatory networks can serve as an important tool for understanding how specific genetic perturbations can affect cell fate and responses. Stochastic models are frequently utilized to capture the random fluctuations of gene expression that have been observed. However, these models tend to be computationally expensive to simulate, and relating them directly to biological measurements is difficult. In this work we introduced new methods for addressing the computational challenges of stochastic gene network modeling. Furthermore, we developed new techniques for model-aided analysis of single cell transcriptomic data.

Specifically, we developed a high-throughput pipeline for computing steady-state solutions of stochastic chemical kinetics models of gene networks. This pipeline enables a comprehensive in silico investigation of gene pair coexpression distributions. Next, we introduced a set of bivariate statistical measures that relate coexpression distributions derived from single cell transcriptomic data to the computational models. A major result of this work is that the noisy gene-gene coexpression distributions obtained from single-molecule-resolution experiments contain information about the underlying gene regulatory network’s interactions. In the future, these results and computational tools can be used to improve gene regulatory network inference algorithms by providing a framework for harnessing the information encoded in the stochastic, “noisy” fluctuations in gene expression.