MCPNet

MCPNet reconstructs genome-scale gene networks from gene expression profiles by computing maximum capacity path (MCP) scores to quantify direct and indirect gene–gene interactions.

Key Features:

• Novel Metric - MCP Score: Computes a maximum capacity path (MCP) score that quantifies the relative strengths of direct and indirect gene–gene interactions and distinguishes interactions beyond Pearson correlation.
• Parallelized Framework: Implements an efficient parallelization strategy optimized for large-scale datasets (tens of thousands of genes) and execution across hundreds of CPU cores.
• Unsupervised and Ensemble Approaches: Supports unsupervised network reconstruction and ensemble methods to combine multiple inference strategies.
• High-Quality Network Reconstruction: Produces reconstructed networks with improved accuracy as measured by the Area Under the Precision-Recall Curve (AUPRC) on benchmark evaluations.
• Scalability and Speed: Scales with increasing dataset sizes while maintaining high computational speed and outperforms existing network reconstruction software in speed and scalability.

Scientific Applications:

• Synthetic benchmarks: Validated on synthetic datasets used for gene network reconstruction benchmarks.
• Saccharomyces cerevisiae: Applied to Saccharomyces cerevisiae gene expression data for reconstruction of gene regulatory networks.
• Arabidopsis thaliana: Applied to Arabidopsis thaliana gene expression data for reconstruction of plant gene regulatory networks.

Methodology:

Integration of the maximum capacity path (MCP) score within a parallelized computational framework and support for unsupervised and ensemble network reconstruction methods.