Figure 1. Multicellular factor analysis on cross-condition single-cell data.

Cross-condition single-cell omics data sample the variability of cells across cell types, patients, and conditions. The information of these datasets can be summarized as a multi-view representation, i.e., a collection of matrices containing cell-type features across samples. Multicellular factor analysis repurposes multi-omics factor analysis (MOFA) to simultaneously decompose the variability of multiple cell types and create a latent space that recovers multicellular transcriptional programs. Throughout this manuscript, several applications are presented to show how this analysis can be used for an unsupervised analysis of single-cell data of multiple samples and conditions, for the identification of multicellular disease processes using the inferred latent space, and for a combined analysis of multiple studies across technologies, such as bulk or spatial transcriptomics. Multicellular factor analysis allows for the inclusion of structural or communication tissue-level views in the inference of multicellular programs, and the joint modeling of independent studies. Moreover, projection of new samples into an inferred multicellular space is also possible.