MOFA

MOFA provides a general, unsupervised framework to integrate multi-omics datasets and to uncover principal sources of biological and technical variation as interpretable latent factors. Conceptually, MOFA extends principal component analysis to multiple data “views” (e.g. transcriptomics, proteomics, epigenomics) and supports complex study designs and multiple sample groups.
The current implementation consists of the MOFA2 R package with a Python core (mofapy2) and extended functionality in MOFA+ for single-cell multi-modal data.

Key Features

  • Multi-view integration: Jointly models several omics layers, capturing both shared and view-specific sources of variation in a unified latent factor model.

  • Interpretable latent factors: Factors summarize global sample structure, while feature loadings identify the genes or molecular features that drive each factor.

  • Groups and complex designs: Supports multiple sample groups, batch effects, and continuous covariates such as time or space through the MEFISTO framework within MOFA2.

  • Single-cell and multi-modal support (MOFA+): Scalable analysis of single-cell multi-omics datasets, with extensive documentation and tutorials.

  • Practical ecosystem: R interface (MOFA2) for training and analysis, Python core (mofapy2) for advanced workflows, and interoperability with both Bioconductor and scverse data structures (e.g. SummarizedExperiment, AnnData/MuData).

  • Downstream analysis: Provides explained-variance summaries, factor visualization, clustering in latent space, and convenient export options for further analysis.

Intended Users

MOFA is designed for researchers in bioinformatics, systems biology, and multi-omics data integration who work with heterogeneous, high-dimensional molecular datasets across multiple omic layers (bulk or single-cell).

It is particularly useful for users who want to:

  • Identify latent biological factors that explain variation across omics layers,

  • Integrate multiple data views and sample groups in an unsupervised manner,

  • Achieve interpretable and transparent results, and

  • Use a scalable framework compatible with both R and Python ecosystems.

    Website
    This email address is being protected from spambots. You need JavaScript enabled to view it.

    Neither development nor maintenance funded by de.NBI