Single-cell multi-omics integration

Integrate paired scRNA-seq and scATAC-seq using MOFA or GLUE to create a unified embedding. MOFA identifies shared latent factors across modalities; GLUE performs direct pairing. Both enable joint analysis of gene expression and chromatin accessibility in the same cells.

Use SIMBA for batch integration to remove technical variations across datasets while preserving biological signal. SIMBA performs multi-batch embedding, aligning cells from different studies into a shared coordinate space for comparative analysis.

Yes, use TOSICA for reference transfer to annotate new datasets based on established cell types. TOSICA maps cells from query data to reference annotations, enabling rapid labeling without manual curation.

Apply StaVIA for trajectory cartography to map developmental paths using integrated multi-omics embeddings. StaVIA constructs cell state transitions across RNA and chromatin modalities, revealing trajectory topology and branch points.

Prepare data with scanpy and anndata: normalize gene expression, process ATAC peaks, handle missing values, and ensure consistent cell identifiers across modalities. Proper data formatting prevents integration failures and improves factor quality.

GPU acceleration is recommended for large datasets, especially with MOFA and scVI-tools components. CPU-only processing works for smaller studies but integration scales significantly faster with torch-compatible GPU resources.