How do you integrate multiple samples for Xenium analysis?

Question: How do you integrate multiple samples for Xenium analysis?

Answer:

Currently, 10x Genomics software does not offer a direct, built-in workflow for integrating multiple Xenium samples. However, Xenium data can be integrated by working with the cell-feature matrices in third-party tools, similar to how you would with Chromium or Visium data. Here are the general steps:

• Experimental design:
  
  • Before starting any Xenium analyzer runs, planning your experiment to minimize batch effects can make integration easier. It's best practice to use the same gene panels and same major versions of Xenium Onboard Analysis across your experiments.
• Merging Datasets:
  • Use tools like Seurat's merge() or Scanpy's concat() to combine data matrices.
  • Address potential cell ID conflicts by changing numerical suffixes in the cell IDs to indicate which dataset it originated from (e.g., ffkpbaba-1 becomes ffkpbaba-2) before merging.
  • Since each ROI has its own coordinate space, prevent spatial mixing of cells from different ROIs by adding a large, fixed offset to cell centroid coordinates when merging.
• Normalizing Data:
  • Normalize data before integration. Both LogNormalize and SCTransform have been successfully used with Xenium data, though a formal 10x comparison has not been done.
• Batch Correcting Data (If Needed):
  • If you've used the same gene panel and see no significant batch effects in technical replicates, batch correction is not necessary.
  • If batch correction is required, Harmony is a widely used method for single-cell and spatial data. Other methods such as scVI may work better depending on your data; experiment and evaluate batch effects.
  • Distributing biological and technical replicates across slides can help determine the best batch correction method.

Working with Spatial Information

How you handle the spatial component depends on your analysis goals.

• Region Selection:
  • For comparing gene expression in specific anatomical regions, use Xenium Explorer's lasso tool to select cells in each tissue section and annotate them in the merged data matrix.
  • Alternatively, select cells based on clusters that correspond to your region of interest.
• Aligning Serial Tissue Sections:
  • For 3D tissue mapping or complex analyses with many samples and regions of interest, you might need to align tissues to a common coordinate space.
  • Consider exploring alignment methods in this paper benchmarking various alignment and integration methods.
  • For alignment, consider these tools:
    • STalign: Aligns using the cell centroid locations and and has a well-documented Xenium tutorial: https://jef.works/STalign/notebooks/xenium-xenium-alignment.html 
    • SANTO: A newer method designed for image-based spatial data, with benchmarking: https://www.nature.com/articles/s41467-024-50308-x