Every spatial transcriptomics experiment is a negotiation between two things that cannot both be maximised freely: where a transcript is and how precisely you can say which cell it came from. Bulk sequencing averages away position; conventional single-cell sequencing throws away position entirely. Spatial transcriptomics (ST) keeps the tissue section intact and records gene expression at known coordinates, but the field has split into families of methods that make very different trade-offs. Two that are frequently compared — and frequently confused — are stRNA-seq (Spatial Transcriptomics RNA sequencing, the barcoded-capture family exemplified by array-based methods such as 10x Visium and the original Stähl et al. approach) and Stereo-seq (Spatial Enhanced Resolution Omics sequencing).
Both measure gene expression while preserving spatial context. They differ in how the coordinate system is built, what resolution that system can support, and what that means for the biology you can actually claim. This note compares them on principle, resolution, workflow, and appropriate use, and ends with a practical decision rule.
The core idea: two ways to attach a coordinate to a transcript
All sequencing-based ST methods work by attaching a known spatial barcode to captured mRNA, then reading that barcode alongside the transcript. The barcode is the coordinate. What differs is how the barcodes are laid down and how densely they can be packed.
stRNA-seq (the barcoded-capture family) works the way the first ST method did: a glass slide or array carries a grid of poly(dT)- or probe-based capture spots, each tagged with a known spatial barcode 1. A tissue section is placed on the array, mRNA is captured in place, and sequencing reconstructs an expression map at the resolution of the capture spots — historically tens of micrometres, effectively one or a few cells per spot. Later developments (Slide-seq, Slide-seqV2) replaced the printed array with a lawn of randomly barcoded beads, pushing resolution toward the single-cell scale 2,3, but the organising principle is the same: pre-made barcodes, captured RNA, position inferred from which barcode caught it.
Stereo-seq takes a different route. Rather than printing or settling barcodes onto a surface, it uses DNA nanoballs (DNBs) patterned by lithography into an ultra-dense, ordered array 4. On this array, the barcoded features are so small and so closely packed that the effective resolution reaches the subcellular scale (on the order of hundreds of nanometres to a few hundred nanometres per feature in early implementations, with centre-to-centre spacing on the order of 0.5 µm cited for the DNB grid). Because every feature carries both a coordinate barcode and a molecular identifier, the array can resolve expression within a single cell and even within subcellular structures.
Principle: nanoball arrays vs. capture barcodes
The mechanistic difference has consequences that show up later in the analysis.
stRNA-seq methods prepare the barcode layer first: either a printed array of spotted features, or a bead lawn, or a microfluidic/deterministic barcoding element 1,2. Capture efficiency and feature size set the practical resolution floor. When the feature is larger than a cell, each spot reports a local mixture, and separating that mixture into cell types becomes a computational deconvolution problem rather than a measurement. This is the central limitation of the classic barcoded-capture approach: the grid is coarse relative to the cell.
Stereo-seq fabricates an ultra-high-density array of DNBs — rolling-circle-amplified DNA clusters patterned by photolithography — in which each DNB carries a positional barcode and a unique molecular identifier (UMI) 4. Tissue is placed on the chip, RNA is captured and reverse-transcribed in situ, and the library is read on a sequencer. Because the grid is dense and ordered, the coordinate of every captured molecule can be assigned with very fine granularity. The design was demonstrated in mouse organogenesis, producing an atlas that resolved developing tissues at near-subcellular scale, and in axolotl brain regeneration, where single-cell Stereo-seq linked progenitor states to spatial niches 4,5.
Resolution: subcellular vs. tissue-scale
stRNA-seq in its classic form operates at 10–100 µm — one spot is often several cells, so the readout is a local average 1. Bead-based variants shrink this considerably and approach single-cell resolution 2,3, but even then, a spot is at best "one cell plus a little of its neighbours." The honest framing is that stRNA-seq answers questions at the level of a tissue neighbourhood, while Stereo-seq answers questions at the level of a cell and its interior.
The most quoted difference is resolution, and it is real. Stereo-seq's dense DNB array resolves at the subcellular level (features on the order of 0.5 µm in the original design), which is finer than most sequencing-based ST methods and comparable in ambition to imaging-based approaches such as MERFISH or seqFISH 4,6. At that scale you can begin to ask where within a cell a transcript sits, how transcripts are partitioned between a soma and a process, or how tightly a signalling factor is confined to a niche.
Resolution is not free. Finer features capture fewer transcripts each, so single-feature counts are noisier and the analysis relies more heavily on aggregation — binning neighbouring features, smoothing, or borrowing strength across the grid 6. A subcellular grid does not automatically give subcellular certainty; it gives subcellular measurement units that the analyst must handle carefully.
Data capture, amplification, and tissue integrity
Both families capture RNA in situ and read it by sequencing, so both preserve the spatial relationship between transcript and position — that is the whole point of the technology 7. Where they diverge is in how much the chemistry and the grid perturb the underlying distribution.
stRNA-seq (probe- or barcode-based) can, depending on the platform, be limited by capture efficiency and by the geometry of the barcode features, so some spatial information is smoothed or lost, particularly when features span multiple cells 1. Stereo-seq, by contrast, captures RNA directly on the chip and reads out position from the DNB coordinate; because the barcoding is spatial rather than probe-defined, the method does not depend on a fixed panel of gene-specific probes, and it aims to preserve the original RNA distribution for whole-transcriptome work 4.
The practical caveat, true for both, is that "preserving tissue integrity" and "measuring RNA" compete. Fixation, permeabilisation, and sectioning all shape what is captured, and a clean-looking expression map can still reflect a compromised section. Neither technology removes the need for careful histology and quality control.
Comparison at a glance
| Feature | stRNA-seq (barcoded capture) | Stereo-seq |
|---|---|---|
| Principle | Pre-made spatial barcodes (spotted array, bead lawn, or microfluidic element) capture RNA in place | DNA nanoball (DNB) array patterned by lithography; position from ultra-dense ordered grid |
| Resolution | Tens of micrometres classically; single-cell scale for bead-based variants | Subcellular (~0.5 µm feature scale) |
| Capture | Barcode/probe-dependent capture; mixture per spot when the spot spans several cells | Direct in-situ capture on the chip; full-transcriptome, no fixed gene panel |
| Tissue integrity | Spatial information smoothed or partially lost as feature size grows | Aims to preserve original RNA distribution; ordered grid limits positional ambiguity |
| Best at | Large tissue sections; broad expression maps; cost-sensitive surveys | Subcellular and single-cell structure; niches; whole-transcriptome at fine scale |
| Main cost | Resolution limits; deconvolution rather than direct cell-level measurement | Sparser per-feature counts; heavier aggregation and QC |
Application scenarios
Mapping the same two methods onto common research questions makes the trade-offs concrete. The following scores summarise practical suitability (★★★★★ = strongest fit):
| Application scenario | stRNA-seq | Stereo-seq |
|---|---|---|
| Large-scale tissue (whole embryo, brain) | ★★★★ | ★★★★ |
| Single-cell & subcellular structure (e.g. synapses) | ★★★ | ★★★★★ |
| Preservation of tissue integrity / RNA distribution | ★★★ | ★★★★★ |
| Gene-expression precision | ★★★ | ★★★★★ |
| Immune & tumour disease research | ★★★★ | ★★★★★ |
Large-scale tissue surveys. Both are capable. If the question is "what does this whole organ or tumour section look like, and where are the major domains," a barcoded-capture stRNA-seq run is often the economical choice, and bead-based variants extend it toward single-cell granularity 2,3,7.
Single-cell and subcellular questions. This is where Stereo-seq's dense array earns its keep — resolving intracellular structure, thin processes, or tightly localised signalling is not something a multi-cell spot can do 4. The axolotl brain-regeneration study is a good illustration of what single-cell Stereo-seq enables when a progenitor population and its spatial niche are the object of study 5.
Immune and tumour microenvironments. Both apply. Stereo-seq's finer scale suits questions about how immune cells are arranged relative to tumour boundaries; stRNA-seq's broader coverage suits mapping the overall architecture and then zooming in computationally.
Developmental and neurobiology work. When cell identity and position together determine the biology — as in organogenesis or regeneration — the higher-resolution method gives a more direct answer, and the resulting atlas can be reused as a reference 4.
How the data are handed to the analyst
Whichever platform produces it, the output is a count matrix plus a coordinate map, and the downstream work looks similar: quality control, normalisation, clustering with spatial awareness, domain detection, and cell-type deconvolution where resolution demands it. Toolkits such as Giotto and the general spatial-analysis ecosystem are designed to handle both array-based and bead-based data 8,6. The difference that matters at this stage is the count density per feature: a subcellular grid hands you many sparse features and expects you to aggregate; a coarse grid hands you fewer, richer spots and expects you to deconvolve. Both are compromises against the same ideal — one transcript, one known cell, no ambiguity — which no current method fully reaches 7.
Practical decision rule
- If the question needs single-cell or subcellular precision — intracellular localisation, fine niches, developmental or neural architecture — prefer Stereo-seq.
- If the question is about large tissue sections and broad expression domains, where neighbourhood-level resolution is enough, stRNA-seq (including bead-based variants) is often the more economical choice.
- If the budget is tight and the biology is tissue-scale, start with the coarser method and use deconvolution to recover cell-type composition.
- Whichever you choose, budget for the analysis cost of the resolution you bought: a finer grid is not automatically a cleaner answer, and a coarser grid is not automatically unusable — it just moves the hard part from measurement to inference.
Conclusion
stRNA-seq and Stereo-seq are both sequencing-based spatial transcriptomics technologies, and both keep position while measuring expression. They are not competitors so much as instruments with different resolving power. stRNA-seq's barcoded capture is a workhorse for tissue-scale maps, and its bead-based descendants narrow the resolution gap at lower cost. Stereo-seq's DNA-nanoball array pushes resolution to the subcellular scale, which is decisive when cell identity and subcellular structure are the biology. The right question is never "which is better" but "which resolution does my question actually require" — and then paying, honestly, for the analysis that resolution demands 6,7.
References
- Stähl PL, Salmén F, Vickovic S, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78–82.
- Rodriques SG, Stickels RR, Goeva A, et al. Slide-seq: A scalable technology for measuring genome-wide expression at high spatial resolution. Science. 2019;363(6434):1463–1467.
- Stickels RR, Murray E, Kumar P, et al. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2. Nat Biotechnol. 2021;39(3):313–319.
- Chen A, Liao S, Cheng M, et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays. Cell. 2022;185(10):1777–1792.e21.
- Wei X, Fu S, Li H, et al. Single-cell Stereo-seq reveals induced progenitor cells involved in axolotl brain regeneration. Science. 2022;377(6610):eabp9444.
- Moses L, Pachter L. Museum of spatial transcriptomics. Nat Methods. 2022;19(5):534–546.
- Rao A, Barkley D, França GS, Yanai I. Exploring tissue architecture using spatial transcriptomics. Nature. 2021;596(7871):211–220.
- Dries R, Zhu Q, Dong R, et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biol. 2021;22(1):78.