Sequencing-Based Spatial Transcriptomics¶
Sequencing-based spatial methods capture the full transcriptome without requiring a pre-selected gene panel. They achieve this by spatially barcoding mRNA molecules on a tissue section, then reading out gene identity and location via next-generation sequencing. The trade-off: most methods capture at multi-cell or bin-level resolution, making computational deconvolution or bin-to-cell aggregation a necessary step.
Visium¶
How it works: Tissue is placed on a slide printed with ~5,000 barcoded spots (55 um diameter, 100 um center-to-center), each capturing mRNA from the overlying cells via poly-dT oligos.
| Property | Value |
|---|---|
| Resolution | 55 um (~1-10 cells per spot) |
| Throughput | Whole transcriptome (~20,000 genes) |
| Tissue area | 6.5 mm x 6.5 mm capture area |
| Tissue type | Fresh-frozen; FFPE via CytAssist |
| Commercial status | Active (10x Genomics). Being superseded by Visium HD but still widely used. |
| Key paper | Stahl et al., Science 2016; 10x Visium protocol |
Recommended analysis tools
- Pre-processing: SpaceRanger, SpotClean (ambient RNA correction)
- Spatial domains: BayesSpace, STAGATE
- Deconvolution: Cell2location, RCTD, Tangram
- SVG detection: nnSVG, SPARK-X
Limitations: 55-um spots mix multiple cells, requiring deconvolution. Limited capture area restricts whole-organ studies. Sensitivity (~3,000-5,000 genes per spot) is lower than scRNA-seq.
Visium HD¶
How it works: Same capture chemistry as Visium but printed at 2-um resolution (continuous lawn of barcoded oligos), with counts aggregated into 2-um or 8-um bins computationally.
| Property | Value |
|---|---|
| Resolution | 2 um bins (approaching single-cell) |
| Throughput | Whole transcriptome |
| Tissue area | 6.5 mm x 6.5 mm |
| Tissue type | Fresh-frozen, FFPE |
| Commercial status | Active (10x Genomics, launched 2024) |
| Key paper | 10x Genomics Visium HD; Oliveira et al., Nature 2024 |
Recommended analysis tools
- Bin-to-cell: Bin2Cell, ENACT
- Spatial domains: BANKSY, STAGATE (must scale to millions of observations)
- Deconvolution: Cell2location, RCTD (on 8-um bins if not using Bin2Cell)
- See the Visium HD deep read for detailed guidance
Limitations: Raw 2-um bins are not cells -- aggregation is essential. Datasets are 10-100x larger than Visium, stressing memory and compute. Tooling ecosystem is still maturing. Per-bin sensitivity is low; aggregation improves it at the cost of resolution.
Slide-seq / V2¶
How it works: A "puck" coated with DNA-barcoded beads (10 um diameter) is placed on a tissue section. mRNA transfers to beads, is captured, and sequenced. Bead positions are decoded by in situ sequencing.
| Property | Value |
|---|---|
| Resolution | 10 um (~1-2 cells per bead) |
| Throughput | Whole transcriptome |
| Tissue area | 3 mm diameter puck |
| Tissue type | Fresh-frozen only |
| Commercial status | Academic protocol (not commercially available) |
| Key paper | Rodriques et al., Science 2019; Stickels et al., Nat Biotechnol 2021 (V2) |
Recommended analysis tools
- Deconvolution: RCTD (originally developed for Slide-seq), Cell2location
- Spatial domains: BayesSpace, STAGATE
Limitations: Fresh-frozen only. Lower sensitivity than Visium per bead (V2 improved ~10x over V1). Small capture area limits applications. Requires custom bead synthesis and decoding.
Stereo-seq¶
How it works: DNA nanoballs (DNBs) are patterned on a chip at ~500 nm spacing, creating a dense grid of spatial barcodes. Tissue is placed on the chip and captured mRNA is sequenced.
| Property | Value |
|---|---|
| Resolution | ~500 nm (subcellular) |
| Throughput | Whole transcriptome |
| Tissue area | Up to several centimeters |
| Tissue type | Fresh-frozen |
| Commercial status | Active (BGI/STOmics) |
| Key paper | Chen et al., Cell 2022 |
Recommended analysis tools
- Initial processing: SAW (BGI pipeline)
- Spatial domains: STAGATE, BANKSY (must handle extreme scale)
- Deconvolution: Cell2location (on aggregated bins)
Limitations: Generates the largest spatial datasets of any technology -- a single sample can produce hundreds of millions of data points. Most standard tools fail at this scale without subsampling. BGI ecosystem is less integrated with the Western scverse/Bioconductor toolchains. Fresh-frozen only.
Open-ST¶
How it works: Open-source protocol using commercially available reagents to create spatial barcoding arrays at subcellular resolution. Designed for accessibility -- any lab with standard molecular biology equipment can implement it.
| Property | Value |
|---|---|
| Resolution | Subcellular (~0.5-1 um) |
| Throughput | Whole transcriptome |
| Tissue area | Variable (depends on array) |
| Tissue type | Fresh-frozen |
| Commercial status | Open-source (Rajewsky lab) |
| Key paper | Schott et al., Cell 2024 |
Open-ST significance
Open-ST democratizes high-resolution spatial transcriptomics by removing the need for expensive commercial platforms. The trade-off is more hands-on protocol optimization and less turnkey support. Analysis follows patterns similar to Stereo-seq.
Seq-Scope¶
How it works: Repurposes Illumina sequencing flow cells as spatial barcoding arrays, achieving subcellular resolution by leveraging the existing cluster positions on the flow cell as spatial barcodes.
| Property | Value |
|---|---|
| Resolution | Subcellular (~0.5-1 um) |
| Throughput | Whole transcriptome |
| Tissue area | Flow cell tile area |
| Tissue type | Fresh-frozen |
| Commercial status | Academic protocol |
| Key paper | Cho et al., Cell 2021 |
Limitations: Requires access to an Illumina sequencing platform for array preparation. Protocol is technically demanding. Limited adoption outside the originating lab.
DBiT-seq¶
How it works: Deterministic barcoding in tissue using two sets of microfluidic channels (horizontal and vertical) that deliver DNA barcodes to create a grid of unique spatial addresses on the tissue.
| Property | Value |
|---|---|
| Resolution | 10 um or 25 um (determined by channel width) |
| Throughput | Whole transcriptome; also supports ATAC-seq and protein |
| Tissue area | Variable (depends on microfluidic device) |
| Tissue type | Fresh-frozen, FFPE |
| Commercial status | Academic protocol (Fan lab, Yale) |
| Key paper | Liu et al., Cell 2020 |
Multi-omic capability
DBiT-seq is notable as one of the earliest spatial multi-omic methods, enabling simultaneous measurement of transcriptome and chromatin accessibility (or protein) from the same tissue section. See Spatial Multi-Omics.
HDST¶
How it works: High-definition spatial transcriptomics using a bead array with 2-um beads, achieving the highest resolution of early bead-based methods.
| Property | Value |
|---|---|
| Resolution | 2 um |
| Throughput | Whole transcriptome |
| Tissue type | Fresh-frozen |
| Commercial status | Academic proof-of-concept (not widely adopted) |
| Key paper | Vickovic et al., Nat Methods 2019 |
Limitations: Very low capture efficiency limited practical utility. Served as an important proof-of-concept for high-resolution sequencing-based spatial transcriptomics but was superseded by Slide-seq V2, Stereo-seq, and Visium HD.
Pixel-seq¶
How it works: Uses polony (polymerase colony) technology to create spatially barcoded arrays at subcellular resolution, with improved transcript capture compared to bead-based approaches.
| Property | Value |
|---|---|
| Resolution | Subcellular |
| Throughput | Whole transcriptome |
| Tissue type | Fresh-frozen |
| Commercial status | Academic protocol |
| Key paper | Fu et al., Cell 2022 |
Limitations: Limited adoption. The polony-based approach is technically demanding to reproduce. Proof-of-concept data looks promising but independent validation is sparse.
Summary comparison¶
| Technology | Resolution | Tissue area | FFPE? | Commercial? | Maturity |
|---|---|---|---|---|---|
| Visium | 55 um | 6.5 mm | Yes | Yes | High |
| Visium HD | 2 um | 6.5 mm | Yes | Yes | Medium |
| Slide-seq V2 | 10 um | 3 mm | No | No | Medium |
| Stereo-seq | 500 nm | cm-scale | No | Yes | Medium |
| Open-ST | Subcellular | Variable | No | Open-source | Early |
| Seq-Scope | Subcellular | Flow cell | No | No | Early |
| DBiT-seq | 10-25 um | Variable | Yes | No | Early |
| HDST | 2 um | Small | No | No | Superseded |
| Pixel-seq | Subcellular | Variable | No | No | Early |