Research Intelligence
The Diffusion Index — Spatial
We can locate any technology on the adoption curve by measuring what fraction of its papers appear in top-tier journals. Early technologies are so novel that just using them earns a Nature paper. Mature technologies don't. Here's where spatial transcriptomics sits today.
In 2020, 146 out of every 1,000 spatial transcriptomics papers appeared in Nature, Science, or Cell. The field was so new that demonstrating spatial gene expression in tissue was itself the finding. Science named spatial transcriptomics its Method of the Year — the field’s equivalent of a Nobel nomination.
By 2025, that number had fallen to 12.4 per 1,000 — a 12× decline in five years. Spatial transcriptomics is no longer a novelty. It is becoming infrastructure. But unlike NGS, which took 16 years to commoditize, spatial appears to be moving faster — driven by commercial platforms from 10x Genomics, NanoString, and Vizgen compressing the adoption timeline.
We have a precedent for exactly this curve. NGS peaked at 118 per 1,000 in 2008 and hit its commoditized floor of ~0.6 by 2016. Spatial transcriptomics is following the same trajectory, shifted roughly 12 years later — and likely moving faster.
2020
Peak year
146.3 per 1,000
12×
Decline peak→today
2020 → 2025
~12
Current score
per 1,000 — still falling
How many papers, and where they land
Annual spatial transcriptomics papers split into three tiers — Nature / Science / Cell, the rest of the top-tier specialist journals (Tier 1+2), and everything else. The log view keeps all three visible; the widening gap is the dilution the curve below captures as a ratio.
Log scale — each line is a count, so all three tiers stay visible despite spanning four orders of magnitude. The gap between the lines is the dilution: top-3 output barely moves while total volume explodes. 2006–2025 (2026 partial year omitted).
Spatial transcriptomics — top-tier journal share over time
Papers per 1,000 spatial transcriptomics publications appearing in top-tier journals, 2015–2026. Shaded regions correspond to Rogers adoption curve phases. Based on 6,653 PubMed-indexed papers. Early years (faded dots) have small sample sizes and should be read with caution.
Peak: 500.0 per 1,000 in 2015 · current (2025): 99.8 per 1,000 — and still declining
What was hot in Nature, Science & Cell
The 146 spatial papers that reached the top three journals, by research theme. Unlike NGS — where new methods earned the premium — spatial’s top-journal currency is the tissue atlas: mapping where cells sit in developing and diseased organs, the spatial context that single-cell sequencing throws away. And unlike NGS, the premium is still rising — the top journals reward spatial more each year, exactly what a technology still mid-curve should show.
Absolute counts — each block is a real paper. Every spatial transcriptomics paper in Nature, Science, or Cell (2015–2025), by research theme — absolute counts. Themes classified by Claude (Haiku) from titles.
What the curve tells us
The 2020 peak was the “method of the year” premium.
At 146 per 1,000, getting a spatial transcriptomics paper into Nature required little more than applying the technology to an interesting tissue. Science’s Method of the Year designation in 2020 marks the peak of the innovator phase — the moment when every top lab recognized spatial as the next must-have capability. That recognition drove a flood of “first spatial map of X” papers, each of which was publishable in top journals simply for being first.
Commoditization is arriving fast — faster than it did for NGS.
The novelty-premium score dropped from 73 in 2022 to 26 in 2023 — a 3× collapse in a single year. NGS took three years for the equivalent fall (2013–2016). The difference is commercial velocity: 10x Xenium launched in late 2022, CosMx in 2022, MERSCOPE in 2021. When instruments ship at scale, the early majority adopts quickly, and the novelty premium evaporates fast. Spatial is compressing roughly eight years of NGS commoditization into about three.
The floor is not yet visible — but we can estimate it.
NGS settled at a floor of ~0.6 per 1,000 by 2016. Spatial is currently at ~12. If spatial follows a similar trajectory, the commoditized floor is probably 3–8 years away — call it 2028–2033. At that point, running a spatial assay will be as unremarkable as running a PCR. The score won’t reach zero: novel applications of spatial (first spatial atlas of a rare tumor, breakthrough clinical use) will keep a residual premium alive indefinitely.
Cost is a brake NGS never had.
NGS cost per genome fell from $1 billion in 2001 to under $1,000 by 2014 — a trajectory that pulled adoption across every lab budget. Spatial has no equivalent collapse underway. A Visium run costs $500–1,000 per sample; Xenium and CosMx sit at $1,500–3,000 per sample. By comparison, bulk RNA-seq — the NGS workhorse most analogous to spatial in terms of per-sample throughput — was already at $50–200 per sample at a comparable adoption stage. Whole-genome sequencing is still $80–150 today on the best platforms; RNA-seq and targeted panels fell much faster. The cost gap between spatial and its NGS predecessor is real and large. This matters because adoption in the early majority — the large hospital labs, the mid-tier biopharma groups — is heavily cost-gated. The result is a two-speed market: academic and large-pharma labs adopting fast (they can absorb the cost), while clinical lab rollout waits for pricing to fall another 5–10×. BGI’s Stereo-seq and Element-adjacent entrants are pushing on this from the commodity end. If a credible $100/sample spatial workflow emerges by 2027, the commoditization timeline compresses significantly. If not, the floor arrives later and sits higher than NGS’s ~0.6.
What this means for platform companies.
When a technology commoditizes, platform economics shift from novelty to throughput, cost, and workflow integration. Illumina learned this the hard way when NovaSeq collapsed per-Gb pricing and compressed margins across the board. Spatial platform companies — 10x, Bruker/NanoString, Vizgen — are currently in the window where switching costs are still high and installed base matters. That window closes as the technology matures. The companies that win are not the ones who owned the innovator phase; they are the ones who own the workflow when the early majority arrives.
The NGS precedent
NGS
Peak: 118 / 1,000 (2008)
Chasm (EA→EM): 2009
Commoditized: 2014–2016
Floor: ~0.6 / 1,000 (2016+)
Time to commoditize: ~8 years
Spatial transcriptomics
Peak: ~146 / 1,000 (2020)
Chasm: 2022–2024 (ongoing)
Floor: TBD — est. 2028–2033
Time to commoditize: est. 5–8 years
What it was used for
The application mix of all spatial papers, 2018–2025, normalized to 100% per year. Note the split personality versus the top journals above: Nature, Science and Cell reward atlases, but the everyday corpus is dominated by cancer and tumor-microenvironment work and ongoing method development — spatial is still building its toolbox while oncologists put it to work.
Research
Translational
Normalized to 100% per year. Classified by Claude (Haiku) over 6,511 spatial papers into tech-specific buckets. 2018–2025.
Spatial is the most research-heavy platform in the index — about 82% basic research in 2025, even more than single-cell. Cancer and the tumor microenvironment are the largest and fastest-growing application (~34% and climbing), and method development is still a quarter of all output, the signature of a field whose toolkit is not yet settled.
But a clinical wedge is opening. Spatial’s natural route to the clinic is digital and spatial pathology, and that translational slice — clinical, pathology, and drug-discovery work — has grown from essentially zero in 2018 to roughly a sixth of output in 2025. That is about where single-cell sat a few years ago: a discovery platform just beginning to cross into application, exactly as its position on the adoption curve would predict.
Who uses it for what
Spatial’s application mix by region. Click regions to set the baseline and see who drives the cancer and pathology work versus the developmental atlases.
Click a region to add or remove it from the comparison. The Average column and the green/red deviations recompute for whatever set you pick — so you can, say, keep only two regions and see how they differ from each other.
| Application | Average 4 regions | |||||
|---|---|---|---|---|---|---|
| Research | ||||||
| Atlas & developmental biology | 10.0% | 10.6% +7% | 15.2% +53% | 11.6% +16% | 7.1% -28% | 8.8% |
| Cancer & tumor microenvironment | 32.7% | 19.4% -40% | 24.6% -25% | 37.4% +15% | 45.2% +38% | 32.0% |
| Immunology & immune niches | 5.8% | 6.8% +15% | 8.6% +47% | 8.8% +51% | 3.8% -34% | 6.3% |
| Neuroscience | 7.4% | 11.7% +57% | 7.6% +3% | 5.4% -27% | 4.4% -40% | 8.5% |
| Methods & computational | 26.7% | 36.1% +35% | 23.3% -13% | 9.5% -64% | 22.6% -15% | 23.2% |
| Other basic research | 1.3% | 1.4% +9% | 1.6% +22% | 0.0% -100% | 1.2% -10% | 1.3% |
| Translational | ||||||
| Drug discovery / target ID | 5.1% | 3.2% -38% | 5.4% +6% | 5.4% +7% | 6.3% +24% | 4.0% |
| Clinical / pathology / translational | 11.0% | 10.8% -2% | 13.6% +24% | 21.8% +98% | 9.3% -15% | 15.9% |
Each cell shows that application’s share within the region, and below it the deviation from the average of the selected regions (e.g. +49% = 1.49× that average). Greyed columns are excluded from the baseline. Cumulative over 6,281 classified papers with a parseable first-author affiliation; region assigned by keyword match.
Where the field is being built
First-author affiliations reveal something unexpected: China has overtaken the United States as the leading producer of spatial transcriptomics research — and the gap is widening. In 2026, Chinese institutions account for nearly twice as many papers as US institutions. This is not a story about replication; it reflects genuine methodological output, driven by BGI’s Stereo-seq program and heavy national investment in genomics infrastructure.
Share of spatial transcriptomics papers by first-author affiliation, 2019–2026. Parsed from 6,403 affiliations. 2019 has N=15 papers — directionally meaningful but statistically noisy.
Sweden’s outsized presence (2% of global output from a country of 10 million) reflects its role as the field’s birthplace. The Ståhl lab at Karolinska/SciLifeLab published the landmark 2016 Science paper that named the technology. That founding advantage seeded an ecosystem that persists today.
The US still leads in commercial platform development — 10x Genomics, Vizgen, and Akoya are all US-headquartered — but China leads in publication volume and is increasingly prominent in platform development through BGI. The implication for the industry: the early majority of spatial adopters, globally, is likely to be Chinese academic and hospital labs. Platform companies that lack a China strategy are ceding that cohort.
Methodology: 6,653 papers from PubMed matching “spatial transcriptomics” OR “spatially resolved transcriptomics,” 2014–2026. Journal tiers: Tier 1 = Nature, Science, Cell (top3); extended Tier 1+2 includes Nature Methods, Nature Biotechnology, Nature Medicine, Nature Genetics, Nature Communications, Nature Cell Biology, Genome Biology, eLife, Cell Genomics, and other high-impact specialist journals. Diffusion score = top-tier papers / total papers × 1,000. Early years (2015–2019) have fewer than 30 papers and should be interpreted with caution. Data computed via scripts/fetch-spatial.ts + scripts/compute-spatial-diffusion.ts. See also: The NGS Diffusion Index.