ModularBoost

ModularBoost infers gene regulatory networks (GRNs) from transcriptomic data by decomposing gene expression into functional modules to constrain network topology and improve inference accuracy and efficiency.

Key Features:

• Module detection (ICA-based): Employs decomposition-based techniques, particularly Independent Component Analysis (ICA)-based methods, to identify functional gene modules from transcriptomic datasets.
• Transcriptomic data support: Operates on time-series expression data, curated datasets, and single-cell RNA sequencing (scRNA-seq) data.
• Topological constraints: Uses identified gene modules as topological constraints to guide network inference.
• Modular inference: Decomposes GRN inference into separate inference of intra-modular and inter-modular interactions.
• Accuracy and efficiency improvements: Incorporates module decomposition and topological constraints to improve inferred network accuracy and computational efficiency.
• Biophysical relevance and interpretation: Integrates module decomposition to enhance the biophysical relevance of inferred networks and provide clearer biological interpretations.
• Benchmarking: Experimental evaluations report outperforming established GRN inference algorithms, particularly on scRNA-seq datasets.

Scientific Applications:

• GRN reconstruction from transcriptomes: Infers gene regulatory networks from time-series, curated, and single-cell transcriptomic data.
• scRNA-seq network analysis: Improves regulatory inference specifically for single-cell RNA sequencing datasets.
• Temporal regulatory dynamics: Supports analysis of time-series expression to capture temporal aspects of regulation.
• Modular organization studies: Enables investigation of intra-modular and inter-modular regulatory relationships within networks.
• Biophysical interpretation of networks: Facilitates generation of networks with enhanced biophysical relevance for biological interpretation.

Methodology:

Detects gene modules using decomposition-based techniques (particularly ICA-based methods), applies identified modules as topological constraints, and performs separate inference of intra-modular and inter-modular interactions.