Discovery of two-level modular organization from matched genomic data via joint matrix tri-factorization

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Advancements in biotechnology provide multi-dimensional genomic data.
• Studying regulatory associations across multiple molecular levels is crucial.
• Existing tools may not fully capture inter-module relationships.

Purpose of the Study:

• To develop a novel computational framework for integrating and analyzing multi-dimensional genomic data.
• To identify modular patterns and their interconnections across different genomic layers.
• To provide comprehensive insights into molecular regulatory mechanisms.

Main Methods:

• A novel non-negative matrix factorization (NetNMF) framework was developed.
• NetNMF integrates pairwise genomic data in a network manner.
• The method was validated using simulated data and compared against existing NMF techniques.

Main Results:

• NetNMF successfully identified modules and their relationships in both simulated and real-world datasets.
• Application to The Cancer Genome Atlas data revealed a two-level microRNA-gene module network.
• Analysis of Cancer Genome Project data identified a two-level gene-drug module network with significant functional implications.

Conclusions:

• The NetNMF tool effectively reveals multi-level modular patterns and their associations.
• Identified modules and their connections demonstrate significant biological relevance.
• This approach enhances understanding of molecular cooperation in biological systems.