MONTI

MONTI integrates genomics, transcriptomics, proteomics, and other omics into a three-dimensional tensor and applies non-negative tensor decomposition to extract subtype-specific multi-omics features for cancer subtype classification and clinical attribute analysis.

Key Features:

• Multi-Omics Integration: Integrates genomics, transcriptomics, proteomics, and other omics into a three-dimensional tensor that represents multi-layered biological data.
• Non-negative Tensor Decomposition: Applies non-negative tensor decomposition to the integrated tensor to extract informative latent components that support subtype-specific feature identification.
• Feature Selection: Selects subtype-specific multi-omics features and identifies gene sets that are strongly regulated by particular omics layers.
• Improved Classification Accuracy: Utilizes combined multi-omics features to improve cancer subtype classification accuracy relative to single-omics gene expression analyses, demonstrated in breast, colon, and stomach cancer cohorts.
• Clinical Attribute Analysis: Analyzes associations between multi-omics-derived features and clinical attributes across nine cancer types to elucidate molecular–clinical correlations.

Scientific Applications:

• Cancer subtype classification: Provides subtype-specific feature sets and latent components for improved classification of cancer subtypes, with demonstrated performance in breast, colon, and stomach cancer cohorts.
• Biomarker and clinical correlation discovery: Enables identification of candidate gene sets and regulatory mechanisms as potential biomarkers or therapeutic targets and assessment of their correlation with clinical attributes across nine cancer types.

Methodology:

Multi-omics datasets are integrated into a three-dimensional tensor; non-negative tensor decomposition is applied to extract components; subtype-specific features are selected and their regulatory patterns across omics layers are analyzed.