Gene regulatory network modeling via global optimization of high-order dynamic Bayesian network

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Dynamic Bayesian Networks (DBNs) are widely used for modeling biological networks like gene regulatory networks (GRNs).
• Current DBN learning methods often rely on local search or meta-heuristic global optimization, yielding suboptimal solutions and limiting applications to small networks.

Purpose of the Study:

• To develop a deterministic global optimization approach for accurate GRN inference from time-course gene expression data.
• To overcome the limitations of existing methods in terms of solution optimality and scalability.

Main Methods:

• Introduced GlobalMIT+, a novel DBN approach utilizing deterministic global optimization.
• Employed the mutual information test (MIT) as an information-theoretic scoring metric.
• Developed a polynomial-time algorithm for learning optimal network structures with inter-time slice arcs.

Main Results:

• GlobalMIT+ successfully learns high-order, time-delayed genetic interactions.
• Demonstrated significantly improved performance compared to existing techniques on synthetic and real biological datasets.
• Successfully applied to a large-scale dataset of 733 cyanobacterial genes.

Conclusions:

• Deterministic global optimization methods are effective for inferring large-scale genetic networks.
• GlobalMIT+ offers a scalable and accurate solution for GRN reverse engineering.