SAILoR: Structure-Aware Inference of Logic Rules.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Boolean networks are effective for modeling gene regulatory network (GRN) dynamics.
• Inferring accurate Boolean GRNs is challenging due to limited experimental data and information loss during binarization.
• Existing methods often produce overfitted models when relying solely on binarized time-series data.

Purpose of the Study:

• To develop a novel method for inferring accurate Boolean models of GRNs.
• To incorporate prior knowledge of network structure alongside time-series gene expression data.
• To improve the accuracy and biological relevance of inferred Boolean GRNs.

Main Methods:

• Proposed SAILoR (Structure-Aware Inference of Logic Rules), a method combining time-series gene expression data with reference networks.
• SAILoR extracts topological properties from reference networks to guide inference.
• Employed the NSGA-II multi-objective genetic algorithm to balance topological similarity and data correspondence.

Main Results:

• SAILoR infers accurate and biologically relevant Boolean GRN models from both static and dynamic perspectives.
• Demonstrated improved static accuracy compared to the dynGENIE3 method.
• Showcased enhanced structural correctness and maintained dynamic accuracy when incorporating prior network structure knowledge.

Conclusions:

• SAILoR effectively integrates diverse data sources and prior knowledge for robust Boolean network inference.
• The method offers a significant advancement in generating reliable Boolean models of GRNs.
• Applied SAILoR to infer context-specific Boolean subnetworks in Drosophila melanogaster, demonstrating practical applicability.