GlobalMIT

GlobalMIT learns globally optimal Dynamic Bayesian Network (DBN) structures using an information-theoretic Mutual Information Test (MIT) criterion to model biological networks such as gene regulatory networks (GRNs) from gene expression data.

Key Features:

• Information-Theoretic Scoring Metric: Uses the Mutual Information Test (MIT) criterion to evaluate dependencies between variables for DBN structure scoring.
• Global Optimization: Performs global optimization to learn DBN structures in polynomial time, addressing the NP-hard nature of DBN structure learning.
• Implementation: Implemented in Matlab and C++.
• Application to Gene Expression Data: Tailored for analyzing gene expression data for modeling gene regulatory networks.

Scientific Applications:

• Gene Regulatory Networks (GRNs): Infers DBN structures to model regulatory relationships and temporal dependencies among genes from expression data.
• Biological Network Modeling: Learns optimal DBN structures for modeling other biological networks beyond GRNs.

Methodology:

Applies the Mutual Information Test (MIT) as the scoring metric and performs polynomial-time global optimization to identify globally optimal DBN structures.