GeneNetTools

GeneNetTools performs differential network analysis of Gaussian graphical models (GGMs) by correcting Ledoit-Wolf shrinkage effects in partial correlations to infer gene regulatory networks from high-dimensional gene-expression data.

Key Features:

• Confidence Intervals: Provides confidence intervals for partial correlations to assess the reliability of inferred network edges.
• Null-Effect Testing: Implements statistical tests for zero partial correlation to identify non-significant interactions.
• Comparative Analysis: Compares partial correlations across different conditions or datasets for differential network analysis.
• Shrinkage Correction: Accounts for shrinkage applied to partial correlations using statistical properties derived from Ledoit-Wolf shrinkage.
• Accounting for Dimensionality and Sample Size: Explicitly incorporates the number of variables, sample size, and shrinkage values to reduce bias in estimates.
• Performance and Validation: Demonstrates computational efficiency and, in simulations, a superior balance of true and false positives compared with DiffNetFDR.
• Software Format: Provided as an R package.

Scientific Applications:

• Gene regulatory network reconstruction: Reconstructs gene regulatory networks from high-dimensional gene-expression profiles.
• Differential network analysis: Detects condition-specific changes in partial correlations across datasets or experimental conditions.
• Validation and benchmarking: Validated using synthetic data and gene-expression datasets from Escherichia coli and Mus musculus.

Methodology:

Applies Ledoit-Wolf shrinkage and corrects for its effects on partial correlations; computes confidence intervals and tests for zero partial correlation; compares partial correlations across conditions while accounting for number of variables, sample size, and shrinkage values; and uses simulations for performance evaluation.