Multi-Omics Factor Analysis-a framework for unsupervised integration of multi-omics data sets.

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Multi-omics studies offer comprehensive biological insights but lack robust unsupervised integration methods.
• Heterogeneous data from different molecular layers present challenges for analysis.

Purpose of the Study:

• To introduce Multi-Omics Factor Analysis (MOFA), a novel computational approach for unsupervised integration of multi-omics data.
• To identify principal sources of variation and disentangle biological and technical heterogeneity across data modalities.

Main Methods:

• Developed MOFA, a factor analysis model to infer latent factors from heterogeneous multi-omics datasets.
• Applied MOFA to chronic lymphocytic leukemia (CLL) patient data (somatic mutations, RNA expression, DNA methylation, drug response).
• Utilized MOFA for single-cell multi-omics data analysis.

Main Results:

• MOFA successfully identified major dimensions of heterogeneity in CLL, including known factors (IGHV status, trisomy 12) and novel drivers (oxidative stress response).
• The method effectively disentangled shared and modality-specific sources of variation.
• MOFA enabled downstream analyses such as sample subgroup identification and outlier detection.

Conclusions:

• MOFA provides a powerful framework for unsupervised integration and interpretation of multi-omics data.
• The approach facilitates the discovery of biological insights and disease drivers.
• MOFA is applicable to both bulk and single-cell multi-omics datasets, revealing coordinated molecular changes.