NExUS

NExUS estimates multiple biological networks from high-throughput genomics and proteomics data across heterogeneous sub-populations with unequal sample sizes using a Bayesian joint-precision-matrix approach.

Key Features:

• Bayesian framework: Employs a Bayesian approach to simultaneously estimate precision matrices across different sub-populations.
• Handling unequal sample sizes: Adjusts joint network estimation to mitigate artifacts in sparsity and inference arising from varying sample sizes.
• Simulation and real-data validation: Includes simulation-based evaluation and has been applied to proteomic networks in cancer research.
• Network similarity and shared pathway analysis: Enables assessment of network similarity and identification of shared pathway activities among related groups.
• Partial-correlation estimation and similarity indices: Produces estimated partial correlation matrices and computes pairwise similarity indices (e.g., lambda_2_square_mean) between categories.

Scientific Applications:

• Proteomic network analysis in related cancers: Supports systems-level network analyses for cancer subtypes, including rare sub-types with limited sample sizes.
• Comparative network studies with TCGA and other datasets: Facilitates comparative inference of pathway and network differences using resources such as The Cancer Genome Atlas (TCGA).

Methodology:

Uses a Bayesian estimation of precision matrices across sub-populations; 'Data_generate.m' generates precision matrices and simulated datasets; 'NExUS.m' computes estimated partial correlation matrices (stored in Partial_corr_mean) and a similarity index for category pairs (lambda_2_square_mean); real-data recommendation is to normalize each variable to mean 0 and variance 1.