MONTI: A Multi-Omics Non-negative Tensor Decomposition Framework for Gene-Level Integrative Analysis.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Multi-omics data integration is crucial for understanding complex biological mechanisms and phenotypes.
• High dimensionality and complex relationships in multi-omics data hinder the association of omics features with specific biological traits.
• Current gene expression data poorly classifies clinically valuable cancer subtypes.

Purpose of the Study:

• To develop and demonstrate a multi-omics analysis method, MONTI (Multi-Omics Non-negative Tensor decomposition for Integrative analysis), for selecting trait-specific multi-omics features.
• To enhance cancer subtyping by integrating and analyzing multi-omics data.
• To identify subtype-specific multi-omics features and infer correlations between omics types.

Main Methods:

• Integration of multi-omics data into a three-dimensional tensor.
• Decomposition of the tensor using non-negative tensor decomposition.
• Application of MONTI to breast, colon, and stomach cancer cohorts for subtype classification and feature selection.

Main Results:

• Subtype classification accuracy significantly improved across three cancer cohorts (breast, colon, stomach) when using integrated multi-omics data with MONTI.
• MONTI identified subtype-specific gene sets strongly regulated by certain omics, enabling inference of correlations between omics types.
• Analysis of clinical attributes across nine cancer types revealed that MONTI could effectively explain certain clinical features using multi-omics data.

Conclusions:

• Integrating multi-omics data using a gene-centric approach, as implemented in MONTI, enhances the detection of cancer subtype-specific and clinical features.
• MONTI provides a powerful framework for understanding the molecular characteristics of cancer subtypes.
• The developed method and associated data offer valuable resources for cancer research.