Skip to content

What is Multi-Sample Analysis?

Multi-Sample Analysis: FICTURE combines multiple related samples and learns a shared set of spatial factors across all of them — then decodes each individual sample against that shared factor set. This builds a common "dictionary" of tissue patterns across all slides.

Multi-Sample Analysis is different from single-sample analysis, where FICTURE learns spatial factors from one sample alone, then decodes those factors back to that sample's own pixels or regions.

How to identify a factor layer is from multi-sample analysis?

When browsing factor sets to add to the map (in the Add Layer dialog or catalog list), a yellow MULTI-SAMPLE ANALYSIS chip is shown next to multi-sample factor sets.

What are shared and per-sample results from a multi-sample analysis?

In multi-sample analysis, CartoScope balances shared global signatures with sample-specific local measurements:

What is Shared Globally?

  • Factor Definitions (Gene Signatures): The genes that define each factor are learned globally.
  • Factor Order: The factors in the Factor Drawer's overview table are sorted consistently across all samples by their combined global weight.

What is Computed Per-Sample?

  • Spatial Map Visualization: Where the factors localize on the tissue (computed from each slide's unique coordinates).
  • Factor Weight / Abundance: The overall presence of a factor on the specific active slide.
  • Marker Genes List: Top markers are computed from the active sample's own data, capturing sample-specific expression variations.
  • UMAP Position: The UMAP layout reflects the active sample's unique spatial factor distribution.

Why Use Multi-Sample Analysis?

Multi-sample analysis is especially useful when analyzing multiple sections from the same tissue, biological replicates, or comparative study cohorts:

  • Consistency: Because every sample is decoded against the same factor definitions, Factor 7 represents the exact same biological pattern in every sample, making cross-sample comparisons direct and meaningful.
  • Robustness: Learning factor signatures from multiple samples simultaneously increases statistical power and helps resolve rarer cell-type patterns.

Related Pages

Concepts

Manual Pages

How-to Guides