DeepHisCoM

DeepHisCoM models complex nonlinear and hierarchical relationships between biological factors and pathways using deep learning-based hierarchical structured component models for pathway analysis.

Key Features:

• Nonlinear Relationship Modeling: Uses deep learning to identify nonlinear contributions of genes, proteins, and other biological factors to pathways.
• Hierarchical Structured Models: Constructs multilayered hierarchical structured component models that represent direct and indirect effects within biological systems.
• Enhanced Analytical Power: Simulation studies show superior power in detecting nonlinear pathway effects while maintaining comparable performance for linear effects.
• Multi-omics and SNP Compatibility: Applicable to metabolomic, transcriptomic, metagenomic, and single-nucleotide polymorphism (SNP) datasets.

Scientific Applications:

• Hepatocellular Carcinoma (HCC): Applied to metabolomic, transcriptomic, and metagenomic datasets for HCC, identifying lysine degradation, valine, leucine, and isoleucine biosynthesis, and phenylalanine, tyrosine, and tryptophan metabolism as associated pathways.
• COVID-19: Applied to a SNP dataset for COVID-19, identifying the MAPK signaling pathway, GnRH signaling pathway, hypertrophic cardiomyopathy, and dilated cardiomyopathy as pathways linked to disease severity.

Methodology:

DeepHisCoM employs deep learning algorithms within hierarchical structured component models to model interactions between biological factors and pathways, capturing both linear and nonlinear effects.