Balanced mediated pathway detection in genomic data

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Understanding how genomic elements interact to influence phenotypic traits is a key research area.
• Omics technologies generate large datasets (e.g., DNA methylation, transcriptome profiling) requiring methods to identify functional modules.
• Previous methods for module discovery often rely on sparse matrix decomposition techniques.

Purpose of the Study:

• To extend the concept of functional modules to identify variables influencing an outcome through stepwise mediation.
• To develop efficient methods for discovering balanced mediated functional modules.
• To apply these methods to both simulated and real-world cancer omics data.

Main Methods:

• Development of methods to extend functional modules for stepwise mediation analysis.
• Efficient tuning of a cardinality-based sparse singular value decomposition (SVD) for module discovery.
• Testing the proposed methods on simulated datasets with known signal and non-signal variables.

Main Results:

• The study presents an efficient approach to tune sparse SVD for discovering balanced mediated functional modules.
• The methods are validated on simulated data, demonstrating their capability to identify stepwise functional modules.
• Application to The Cancer Genome Atlas (TCGA) data showcases real-world utility.

Conclusions:

• The developed methods provide an efficient way to discover balanced mediated functional modules from omics data.
• This approach enhances the understanding of how different genomic variables collectively influence phenotypic outcomes via stepwise mediation.
• The findings have implications for identifying complex genetic architectures underlying diseases and traits.