By Lociven · SpatiaBio · July 2026
Xenium and Visium are both spatial transcriptomics platforms, but they answer different questions. Choosing the wrong one for your experiment wastes months and tens of thousands of dollars. This post walks through the real differences — resolution, throughput, gene panel, data format — and tells you which to use for which biological question.
The core difference
Visium measures RNA from tissue sections using spatially barcoded capture spots (~55 µm diameter). Each spot captures transcripts from multiple cells — typically 2–10 depending on tissue density. You get whole-transcriptome coverage across thousands of spots.
Xenium is an in situ platform. It detects individual RNA molecules inside intact cells using fluorescent probes, then reads them with iterative cycles of imaging. Each transcript is localized to a single cell, and cells are segmented computationally. You get single-cell resolution — but only for the genes in your panel (up to ~5,000 currently).
Rule of thumb: If you need to discover new cell types or pathways, use Visium (whole transcriptome). If you already know what cell types to look for and want their exact spatial positions, use Xenium.
Side-by-side comparison
| Feature | Visium HD | Xenium |
|---|---|---|
| Resolution | 2 µm bins (HD) / 55 µm spots (v2) | Single cell (~10–20 µm) |
| Gene coverage | Whole transcriptome (~18,000) | Panel-based (280–5,000 genes) |
| Cell segmentation | Requires separate image + algorithm | Built-in (nuclear + cell boundary) |
| Tissue area | ~6.5 × 6.5 mm (v2) | ~1 × 1 cm per run |
| Run time | ~1 day (library prep + sequencing) | ~1–3 days (imaging cycles) |
| Cost per sample | ~$500–800 reagents + sequencing | ~$1,500–2,500 per slide |
| FFPE compatible | Yes (FFPE kit) | Yes (optimized for FFPE) |
| Best use case | Discovery, whole transcriptome atlas | Cell-type mapping, spatial niches |
The same specs as the table above, arranged so the decision is visible. Left: every platform trades spatial resolution against how much of the transcriptome you measure — the top-right corner (single-cell resolution and whole transcriptome) is still empty. Right: what a single measurement covers, drawn to scale — a 55 µm Visium spot averages 5–20 cells into one profile, while Xenium resolves each cell separately.
When to use Visium
Use Visium when you don't know what you're looking for. It captures the full transcriptome, which means you can cluster spots, identify marker genes, run differential expression between regions, and discover spatial domains you didn't know existed.
Good fits:
- First spatial experiment on a tissue type you haven't profiled before
- Identifying tumor microenvironment zones (stromal, immune, hypoxic)
- Comparing spatial gene expression between conditions (treatment vs control)
- Generating hypotheses for follow-up experiments
Visium HD (2 µm bins) closes most of the resolution gap with Xenium for many tissue types. If your core facility has it, run HD over v2 whenever possible.
When to use Xenium
Use Xenium when cell identity matters more than gene discovery. Because it localizes transcripts to individual cells without deconvolution, you get clean cell-level data — especially important in tissues where cells are tightly packed or interdigitated.
Good fits:
- Tumor-immune interface: exact positions of CD8+ T cells relative to cancer cells
- Validating scRNA-seq findings with spatial context
- High-resolution niche analysis (tertiary lymphoid structures, invasive margin)
- FFPE clinical samples where you need single-cell resolution
Watch out: Xenium panel design matters enormously. If you miss key marker genes in your panel, you cannot recover that information post-hoc. Spend time designing the panel before you run the experiment.
Running both together
The strongest experimental design uses both platforms on serial sections from the same tissue block:
- Run Visium for whole-transcriptome discovery → identify spatial domains, candidate cell types, and key marker genes
- Design a Xenium panel from the Visium data
- Run Xenium on an adjacent section for single-cell resolution validation
This is now a standard approach in high-impact spatial studies. The cost is roughly 3–4× a Visium-only experiment, but the resolution of cell-type assignments is incomparably better.
Analyzing Xenium data with Squidpy
Squidpy handles Xenium data natively. The sq.read.xenium() function loads the Xenium output folder directly into an AnnData object with spatial coordinates and transcript-level data.
import squidpy as sq
# Load Xenium output
adata = sq.read.xenium("path/to/xenium_output/")
# Compute spatial neighbors
sq.gr.spatial_neighbors(adata, coord_type="generic", spatial_key="spatial")
# Neighborhood enrichment
sq.gr.nhood_enrichment(adata, cluster_key="cell_type")
sq.pl.nhood_enrichment(adata, cluster_key="cell_type")
The key difference from Visium: Xenium coordinates are in microns (not pixel units), and the cell-level resolution means you can run co-occurrence and neighborhood enrichment at true single-cell scale rather than spot scale.
Other platforms worth knowing
MERSCOPE (Vizgen) — similar to Xenium, MERFISH-based, strong in neuroscience applications. Up to 500 genes per panel in the standard configuration.
CosMx (NanoString) — RNA + protein in the same experiment, good for immune profiling where protein markers are still the standard.
Slide-seqV2 — 10 µm resolution, whole transcriptome, but lower sensitivity than Visium HD. Used in academic settings with existing infrastructure.
Go deeper
SpatiaBio Pack 3 — Platform Selection Guide
Everything above, fact-checked against current spec sheets in one reference PDF — capture area, gene panel size, cost per sample, and the tradeoffs that actually matter when you're the one signing off on the platform choice.
Get Pack 3 →From the NeoantigenLab sister blog
Working with tumor transcriptomics for neoantigen research?
NeoantigenLab covers neoantigen biology, WES pipelines, pVACseq, and HLA typing for experimental researchers — the biology behind what you're sequencing.
Visit NeoantigenLab →Tags: Xenium, Visium, Visium HD, spatial transcriptomics, 10x Genomics, MERSCOPE, CosMx, single-cell resolution, Squidpy, spatial omics comparison
No comments:
Post a Comment