What types of spatial transcriptomics data does UCASpatial support?
UCASpatial supports various spatial transcriptomics platforms, including 10X Visium, Visium HD, and Slide-seq. It works with both R (Seurat) and Python (AnnData) formats.
How do I choose the right single-cell reference data?
Choose single-cell reference data that:
- Comes from the same tissue type as your spatial data (recommended, but an unpaired reference is also fit)
- Has high-quality cell type annotations
- Contains similar cell types expected in your spatial data
- Has been properly quality-controlled and filtered
What should I do if I get low-quality deconvolution results?
Try the following troubleshooting steps:
- Check if your reference data matches the tissue type
- Check the Quality Control Plot (See Tutorial-Visualization)
- Verify cell type annotations in the reference are accurate
- Consider filtering low-quality genes or spots
- Try different parameters, such as meta.purity, meta.resolution
How long does UCASpatial take to run?
Runtime depends on:
- Number of spots in spatial data (typically scales linearly)
- Number of cell types in reference data
- Number of genes used for deconvolution
- Computational resources available
Typical runtime: 5-30 minutes for standard Visium data on a modern laptop.
Can UCASpatial handle large datasets like Visium HD?
Yes, UCASpatial includes specific functions for Visium HD data. Use UCASpatial_HD_deconv() in R or the standard Python implementation with appropriate memory management.
How do I interpret the cell proportion values?
Cell proportions represent the estimated fraction of each cell type at each spatial location. Values range from 0 to 1, where:
- 0 = cell type not detected
- 1 = spot composed entirely of that cell type
- Intermediate values = mixed cell populations