CBP-JMF

CBP-JMF elucidates complex biological processes (CBPs) underlying disease subtypes by jointly factorizing multi-omics data.

Key Features:

• Joint non-negative matrix tri-factorization framework: CBP-JMF uses joint non-negative matrix tri-factorization to decompose multi-omics datasets into components that represent underlying biological processes.
• Decomposition into samples, features, and processes: The method represents data as matrices for samples, features (genes or proteins), and biological processes to capture relationships among sample groups and CBPs.
• Non-negativity for interpretability: Non-negative factorization constraints produce components with biologically interpretable, non-negative values.
• Implementation in Python: The software implementation is in Python.
• Discovery of disease-subtype CBPs: CBP-JMF is configured to identify CBPs that define and separate sample groups corresponding to disease subtypes.

Scientific Applications:

• Multi-omics integration for disease characterization: Integrates multi-omics data to characterize complex biological processes underlying diseases.
• Disease subtype identification: Discovers CBPs that define and group samples into disease subtypes for classification and stratification.
• Pathway and gene association analysis: Extracts genes from inferred CBPs and compares them to known subtype pathways, exemplified by application to breast cancer where CBP-JMF identified CBPs for four distinct subtypes with gene overlap to known pathways.
• Biomarker and mechanism discovery: Supports molecular pathway analysis and identification of candidate biomarkers associated with disease subtypes for systems biology studies.

Methodology:

Applies joint non-negative matrix tri-factorization to multi-omics data, decomposing matrices into sample, feature (genes or proteins), and biological-process components under non-negative factorization constraints.