Combining kinetic orders for efficient S-System modelling of gene regulatory network

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• S-System models, non-linear differential equation models, are crucial for reconstructing gene regulatory networks from temporal gene expression data.
• Learning numerous parameters in S-System models leads to high computational costs and challenges with invalid network reconstructions due to independent parameter convergence.
• Previous methods improved performance but still required significant computation for larger networks.

Purpose of the Study:

• To address the computational expense and invalid network issues in S-System model parameter optimization.
• To develop a novel method that enhances the accuracy and efficiency of gene regulatory network reconstruction.

Main Methods:

• Introduced a novel penalty term to penalize invalid kinetic orders in S-System models.
• Developed a new parameter, w_ij, combining kinetic parameters (g_ij and h_ij) to handle invalid regulations.
• Integrated these features into the Dynamically Regulated Network Initialization (DRNI) algorithm as a third stage, creating a dynamic penalty system.

Main Results:

• The novel method significantly reduced the number of iterations required for convergence in gene network reconstruction.
• Achieved improved network accuracies, demonstrated by a 300-generation reduction and a 0.05 F-score improvement for a 20-gene network.
• Showcased enhanced performance on DREAM challenge datasets, improving the average area under the ROC curve for 10-gene networks.

Conclusions:

• The developed method effectively addresses the challenge of invalid network reconstructions in S-System modeling.
• This approach systematically eliminates invalid networks, encouraging valid candidate solutions and improving computational efficiency.
• The enhanced DRNI algorithm provides a more accurate and computationally feasible solution for gene regulatory network inference.