ModularBoost: an efficient network inference algorithm based on module decomposition.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference aims to decipher regulatory relationships from expression data.
• Existing methods often overlook GRN's topological characteristics, resulting in networks lacking clear biological interpretation.
• This study addresses this by incorporating data-driven module detection prior to network inference.

Purpose of the Study:

• To enhance the biophysical interpretability of inferred GRNs.
• To develop a method that improves the accuracy and efficiency of GRN inference.
• To leverage gene modules as topological constraints for GRN construction.

Main Methods:

• Employed decomposition-based methods for identifying gene modules from transcriptomic data.
• Utilized Independent Component Analysis (ICA)-decomposition for module detection.
• Developed the ModularBoost method for GRN inference incorporating topological constraints.

Main Results:

• ModularBoost demonstrated superior efficiency and accuracy compared to established methods across various datasets (time-series, curated, scRNA-seq).
• The method effectively identified functional gene modules directly from transcriptomic data.
• ModularBoost outperformed other inference algorithms on single-cell RNA sequencing (scRNA-seq) datasets.

Conclusions:

• GRN inference can be simplified by decomposing it into inferring intra-modular and inter-modular interactions.
• The proposed ModularBoost method improves GRN inference accuracy and efficiency by utilizing identified gene modules as topological constraints.
• The findings suggest a more biologically meaningful approach to constructing GRNs.