MONET: Multi-omic module discovery by omic selection.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Modern experimental biology generates large multi-omic datasets, where multiple genome-wide data types are measured per sample.
• Integrative analysis, particularly sample clustering, is crucial for understanding biological processes and identifying disease subtypes.

Purpose of the Study:

• To introduce MONET (Multi Omic clustering by Non-Exhaustive Types), a novel algorithm for multi-omic sample clustering.
• To address limitations of existing methods by allowing distinct clustering structures across different omics.

Main Methods:

• MONET discovers sample modules where each module utilizes a specific subset of available omics.
• The algorithm was rigorously tested on simulated data, an image dataset, and ten multi-omic cancer datasets from The Cancer Genome Atlas (TCGA).

Main Results:

• MONET demonstrated favorable performance compared to existing multi-omic clustering methods.
• The tool proved robust to missing data, capable of gene clustering, and effective in revealing cell type modules in single-cell data.
• Biological and clinical relevance was confirmed through analysis of Ovarian Serous Cystadenocarcinoma data.

Conclusions:

• MONET offers a unique and valuable approach to multi-omic clustering.
• It provides complementary insights to existing algorithms, advancing the understanding of complex biological systems and disease heterogeneity.