Comparative study of three commonly used continuous deterministic methods for modeling gene regulation networks.

Area of Science:

• Computational biology
• Systems biology
• Bioinformatics

Background:

• Gene-regulatory networks (GRNs) control gene transcription rates through DNA, RNA, and protein interactions.
• Accurate dynamic modeling of GRNs is crucial for biomedical research and development.
• Continuous deterministic modeling approaches are key to constructing these dynamic GRN models.

Purpose of the Study:

• To comprehensively compare three common ordinary differential equation systems for dynamic GRN modeling: S-system (SS), artificial neural networks (ANNs), and general rate law of transcription (GRLOT).
• To evaluate the ability of these methods to replicate reference GRN models' regulatory structure and gene expression dynamics under various conditions.

Main Methods:

• Comparative analysis of S-system (SS), artificial neural networks (ANNs), and general rate law of transcription (GRLOT) methods.
• Evaluation of model robustness and accuracy in replicating reference GRN structures and dynamics.
• Cross-method reverse-engineering experiments using simulated data.

Main Results:

• ANN and GRLOT methods produced robust models, even with parameter deviations.
• SS-based models showed performance loss due to complex power terms, even with close parameter matching.
• All methods accurately reproduced training data from single conditions, but only multi-condition data enabled accurate GRN feature reproduction.

Conclusions:

• Dynamic GRN modeling methods differ significantly in replicating network structure and behavior.
• Method selection for GRN modeling requires careful consideration to avoid biased results.
• Relying on a single modeling method may lead to an incomplete or skewed understanding of GRNs.