BREM-SC

BREM-SC performs joint clustering of paired scRNA-seq and CITE-seq count data to integrate transcriptomic and cell-surface proteomic information at single-cell resolution.

Key Features:

• Bayesian Random Effects Mixture model: Employs a Bayesian Random Effects Mixture framework for simultaneous clustering of transcriptomic and proteomic single-cell data.
• Incorporation of correlation via random effects: Uses random effects to model and account for correlation between RNA and protein modalities rather than assuming independence.
• Direct use of raw droplet-based count data: Processes raw count data directly from droplet-based scRNA-seq and CITE-seq experiments without requiring prior transformation.
• Quantification of clustering uncertainty: Provides probabilistic measures of clustering uncertainty for each cell to assess confidence in cluster assignments.

Scientific Applications:

• Immunology: Enables integrated analysis of gene expression and cell-surface protein markers to resolve immune cell types and states.
• Cell-type, state, and interaction discovery: Facilitates discovery of novel cell types, cellular states, or cell–cell interactions that may be missed by single-modality analyses.

Methodology:

BREM-SC is implemented as a direct extension of the DIMMSC model into a Bayesian Random Effects Mixture framework that incorporates random effects to model correlation between scRNA-seq and CITE-seq count data, and it has been validated by simulation studies and analyses of public and proprietary datasets.