single-cell-multi-omics-integration

Cheat sheet for OmicVerse tutorials covering multi-omic integration, batch integration, reference transfer, and trajectory analysis.

• MOFA for paired scRNA+scATAC integration
• MOFA after GLUE pairing alignment
• SIMBA for multi-batch cell embedding
• TOSICA for reference transfer and pathway prediction
• StaVIA for single-cell trajectory cartography

• Uses preprocessed AnnData inputs
• Supports paired and unpaired multi-omic workflows
• Exports models and factors for reuse
• Includes visualization and result inspection tools
• Requires specific dependencies like pyMOFA, pySIMBA, pyTOSICA, pyVIA
• CPU-bound with optional GPU acceleration for MDE/UMAP
• Supports downstream reuse of trained models and factors
• Provides trajectory and pathway-level interpretation

1. 1.Load preprocessed AnnData and initialize model
2. 2.Run model training and save outputs to HDF5 or project folder
3. 3.Inspect results with built-in plotting and factor analysis
4. 4.Reuse saved models for future annotation or visualization
5. 5.Ensure correct gene sets and normalization before training
6. 6.Use appropriate hardware (CPU/GPU) based on memory and compute needs