Differential equation based reverse-engineering algorithms: Pros and cons

Ordinary Differential Equations (ODEs) represent a deterministic approach to model gene regulatory networks. ODEs can be used to model changes in gene transcription induced by an external perturbation, such as gene overexpression/downregulation or treatment with a drug. Reverse-engineering algorithms based on ODEs require a choice of a functional form describing the effect of a regulator on its target genes. Here, we focused on an ODE-based reverse engineering algorithm named Network Identification by multiple Regression (NIR) which is rooted on the hypothesis that the regulation exerted by one gene (i.e., a TF) on a target gene can be approximated by a linear function, i.e., the transcription rate of the target gene is proportional to the amount of TF. NIR uses steady-state gene expression measurements and requires knowledge of the genes perturbed in each experiment. We showed that even if originally NIR was created for a different purpose, it can be successfully used to infer gene regulation from an integrated genotype and phenotype dataset. Our results provide evidence of the feasibility of applying reverse-engineering algorithms, such as NIR, to infer gene regulatory networks by integrated analysis of genotype and phenotype.