bio-multi-omics-mofa-integration

To integrate metabolomics and transcriptomics data, you can use MOFA2 to identify shared and modality-specific latent factors. This requires aligned feature matrices with common sample identifiers, followed by normalization and feature selection before model training.

Yes, MOFA2 supports single-cell multimodal data workflows through the MOFA+ framework. You can apply unsupervised factor discovery to single-cell data to identify latent factors explaining biological variation across modalities, provided matrices are properly aligned and preprocessed.

Data preparation for multi-omics factor analysis involves loading multiple omics matrices, intersecting samples to ensure common identifiers, transposing to MOFA2 view format, and applying normalization and feature selection. Proper preprocessing ensures reliable latent factor extraction across modalities.

To interpret latent factors from multi-omics integration, extract factor scores and weights from the trained MOFA2 model, compute variance explained per view, visualize factors and weights, and perform gene-set or pathway enrichment analysis on top features to derive biological meaning.

Multi-omics factor analysis can be performed using either MOFA2 in R or mofapy2 in Python. Both implementations support model configuration, variational or stochastic inference training, and model saving for downstream analysis and visualization tasks.

MOFA2 multi-omics integration requires aligned feature matrices with common sample identifiers across all views, meaning datasets without shared samples cannot be directly integrated. Additionally, preprocessing including normalization and feature selection is mandatory for reliable factor discovery.