multivarNetwork

multivarNetwork infers regulatory networks by integrating proteomic and transcriptomic data using multi-attribute Gaussian graphical models to model direct interactions across data modalities.

Key Features:

• Integration of Multiple Data Sources: Integrates proteomic and transcriptomic datasets measured on the same set of variables and individuals for joint network inference.
• Multi-Attribute Gaussian Graphical Models (GGM): Extends traditional Gaussian graphical models into a multivariate multiattribute GGM to represent direct links between biological entities across modalities.
• Neighborhood Selection Procedure: Employs a multiscale neighborhood selection procedure to infer network structure within and across data layers.
• Group-Lasso Penalty: Applies a group-Lasso penalty during inference to select relevant interactions across attributes.
• Consensus Network Construction: Constructs a consensus network by integrating inferred interactions across multiple biological scales.

Scientific Applications:

• Regulatory Network Reconstruction: Reconstructs molecular regulatory networks from integrated multi-omics datasets such as proteomic and transcriptomic data.
• Breast Cancer Network Analysis: Demonstrated on breast cancer datasets to integrate proteomic and transcriptomic information for studying disease mechanisms.

Methodology:

Combine multiple datasets measured on the same variables and individuals; model direct interactions using a multivariate multiattribute Gaussian graphical model; infer network edges via a multiscale neighborhood selection procedure with group-Lasso penalties; and integrate inferred interactions into a consensus network across scales.