BRANENET

BRANENet integrates multi-omics data by embedding graph-based information from multilayer heterogeneous networks into a lower-dimensional space to study regulatory mechanisms across molecular layers such as chromatin accessibility, transcription, RNA maturation, transport, and cellular metabolism.

Key Features:

• Multi-omics integration: Integrates omics data across different molecular layers to enable combined analysis of heterogeneous datasets.
• Multilayer heterogeneous network analysis: Operates on multilayer heterogeneous networks derived from multi-omics measurements.
• Graph embedding: Embeds graph-based information into a lower-dimensional space to represent complex interactions among nodes.
• Random-walk matrix factorization: Leverages random walk information within a matrix factorization framework to learn node embeddings.
• Regulatory scope: Captures interactions relevant to chromatin accessibility, transcription, RNA maturation, transport, and cellular metabolism.
• Empirical evaluation: Evaluated on Saccharomyces cerevisiae using RNA sequencing (RNA-seq) and targeted metabolomics via nuclear magnetic resonance (NMR) in a heat-shock time course at 0, 20, and 120 minutes, identifying differentially expressed biomolecules responsive to heat stress.
• Downstream inference: Supports transcription factor (TF)-target prediction, integrated omics network (ION) inference, and module identification.
• Performance: Demonstrated superior performance compared to existing network integration methods by achieving high prediction scores across a variety of downstream tasks.

Scientific Applications:

• Regulatory mechanism analysis: Study gene expression regulation across chromatin accessibility, transcription, RNA maturation, transport, and metabolism.
• TF-target prediction: Infer transcription factor–target relationships from integrated multi-omics networks.
• Integrated omics network (ION) inference: Reconstruct integrated networks that connect different molecular layers.
• Module identification: Detect modules of coordinated biomolecules within multilayer networks.
• Stress-response analysis in yeast: Analyze Saccharomyces cerevisiae heat-shock responses using RNA-seq and NMR metabolomics to identify differentially expressed biomolecules.

Methodology:

Computes node embeddings for multilayer heterogeneous networks by incorporating random walk information into a matrix factorization framework to embed graph-based multi-omics data into a lower-dimensional space.