Inference of gene regulatory networks with multi-objective cellular genetic algorithm

Area of Science:

• Computational Systems Biology
• Bioinformatics
• Network Inference

Background:

• Reverse engineering biochemical networks is crucial for understanding biological systems.
• Current methods often rely on single-objective evaluation (MSE), limiting optimization.
• Gene Regulatory Network (GRN) inference requires accurate topological structure and parameter estimation.

Purpose of the Study:

• To propose a multi-objective approach for inferring S-System structures of GRNs.
• To optimize GRN topology and kinetic parameters (rate constants, orders) concurrently.
• To avoid the use of arbitrary penalty weights in the model inference process.

Main Methods:

• Developed a multi-objective formulation for GRN inference using Pareto dominance and optimality.
• Adapted a Multi-Objective Cellular Evolutionary Algorithm for parameter learning and topology inference.
• Implemented the approach in a software tool named MONET.

Main Results:

• Evaluated MONET on synthetic and real gene expression datasets (DREAM3, IRMA).
• Demonstrated competitive performance compared to existing methods in reproducing biological behavior.
• Showcased robustness to noise and the ability to provide trade-off solutions.

Conclusions:

• The multi-objective approach offers advantages over conventional single-objective methods for GRN inference.
• MONET effectively infers GRN structures and parameters, providing valuable insights into biological dynamics.
• The method facilitates the exploration of diverse GRN typologies and their associated performance trade-offs.