A Model-Based Tool for the Analysis and Design of Gene Regulatory Networks.

Computational and mathematical models have significantly contributed to the rapid progress in the study of gene regulatory networks (GRN), but researchers still lack a reliable model-based framework for computer-aided analysis and design. Such tool should both reveal the relation between network structure and dynamics and find parameter values and/or constraints that enable the simulated dynamics to reproduce specific behaviors. This paper addresses these issues and proposes a computational framework that facilitates network analysis or design. It follows a modeling cycle that alternates phases of hypothesis testing with parameter space refinement to ultimately propose a network that exhibits specified behaviors with the highest probability. Hypothesis testing is performed via qualitative simulation of GRNs modeled by a class of nonlinear and temporal multiscale ODEs, where regulation functions are expressed by steep sigmoid functions and incompletely known parameter values by order relations only. Parameter space refinement, grounded on a method that considers the intrinsic stochasticity of regulation by expressing network uncertainty with fluctuations in parameter values only, optimizes parameter stochastic values initialized by probability distributions with large variances. The power and ease of our framework is demonstrated by working out a benchmark synthetic network to get a synthetic oscillator.