Author:
Erbe Rossin,Stein-O’Brien Genevieve,Fertig Elana
Abstract
AbstractComputer simulations of cell behaviors and dynamics allow for investigation of aspects of cellular biology with a ground truth that is currently difficult or impossible to generate from experimentally generated profiling data. Here, we present a mechanistic simulation of cell states that models the stochastic interactions of molecules revealing the DNA accessibility, RNA expression, and protein expression state of a simulated cell and how these states evolve over time. By designing each component to correspond to a specific biological molecule or parameter, the simulation becomes highly interpretable. From the simulated cells generated, we explore the importance of parameters such as splicing and degradation rates of genes on RNA and protein expression, demonstrating that perturbing these parameters leads to changes in long term gene and protein expression levels. We observe that the expression levels of corresponding RNA and proteins are not necessarily well correlated and identify mechanistic explanations that may help explain the similar phenomenon that has been observed in real cells. We evaluate whether the RNA data output from the simulation provides sufficient information to reconstruct the underlying regulatory relationships between genes. While predictive relationships can be inferred, direct causal regulatory relationships between genes cannot be reliably distinguished from other predictive relationships between genes arising independently from a direct regulatory mechanism. We observe the same inability to robustly distinguish causal gene regulatory relationships using simulated data from the simpler BoolODE model, suggesting this may be a limitation to the identifiability of network inference.
Publisher
Cold Spring Harbor Laboratory