Perspective 169 views
The unequal geography of spatial transcriptomics
Spatial transcriptomics maps gene expression within tissue, and it stands to change how cell biology, pathology and histology are practised. In January 2021, Nature Methods named it Method of the Year 2020. Fewer than 1% of the spatial transcriptomics papers we analysed had a first author in a low- or lower-middle-income economy. As the field expands, who gets to take part?
A research boom
PubMed papers mentioning spatial transcriptomics grew more than 13-fold from 2021 to 2025. Their share of all PubMed records rose from 12 to 160 per 100,000. By 25 September 2026, the 3,748 papers published in 2026 had already surpassed the previous year's total.
Papers naming spatial transcriptomics, by year
- 2019: 32 papers
- 2020: 89 papers
- 2021: 221 papers
- 2022: 467 papers
- 2023: 915 papers
- 2024: 1,580 papers
- 2025: 3,003 papers
- 2026, to 25 September: 3,748 papers
- A whole year
- 2026, to 25 September
Show the numbers as a table
| Year | Papers |
|---|---|
| 2019 | 32 |
| 2020 | 89 |
| 2021 | 221 |
| 2022 | 467 |
| 2023 | 915 |
| 2024 | 1,580 |
| 2025 | 3,003 |
| 2026, to 25 September | 3,748 |
Where the authors work
We compared spatial transcriptomics papers published from 2021 to 2025 with every paper from the same years mentioning haematoxylin and eosin (H&E) staining or histopathology, and with PubMed in general, from two random samples. We grouped authors' affiliations by the World Bank's income classification.
First authors in low- or lower-middle-income economies accounted for 0.7% of spatial transcriptomics papers, compared with 16.7% of H&E and histopathology papers and about 7% of papers in PubMed in general. The picture is the same when every author counts, as in the chart below: 1.6% of spatial transcriptomics papers had at least one author in these economies, against 18.8% of H&E and histopathology papers.
19 in 100
H&E and histopathology
18.8%, 22,665 of 120,514 papers
First author 16.7%, last author 16.5%
2 in 100
Spatial transcriptomics
1.6%, 89 of 5,672 papers
First author 0.7%, last author 0.7%
Show the numbers as a table
| Author | H&E and histopathology | Spatial transcriptomics |
|---|---|---|
| Any author | 18.8%22,665 of 120,514 | 1.6%89 of 5,672 |
| First author | 16.7%19,858 of 119,191 | 0.7%40 of 5,647 |
| Last author | 16.5%19,353 of 117,244 | 0.7%37 of 5,608 |
Researchers in lower-income economies are far better represented in H&E and histopathology research than in spatial transcriptomics. Virtual spatial transcriptomics could help close that gap, because it starts from the H&E images they already work with.
China and the United States account for 70.8% of the spatial transcriptomics papers by first-author affiliation. First authors were based in 7 low- and lower-middle-income economies, compared with 53 upper-middle- and high-income ones.
India holds 8.6% of H&E and histopathology papers and 0.5% of spatial transcriptomics papers
Share of papers from 2021 to 2025 by the first author's country
Showing 10 of 48: most H&E and histopathology papers
Show the table of 48 countries and economies with at least 20 spatial transcriptomics or 250 H&E and histopathology papers
| Country or economyIncome group | H&E and histopathologyof 119,191 | Spatial transcriptomicsof 5,647 |
|---|---|---|
| ChinaUpper middle income | 29.7%35,347 | 38.5%2,174 |
| United StatesHigh income | 10.2%12,156 | 32.3%1,826 |
| IndiaLower middle income | 8.6%10,307 | 0.5%30 |
| JapanHigh income | 5.0%5,972 | 2.8%160 |
| TürkiyeUpper middle income | 4.2%5,047 | 0.1%6 |
| Egypt, Arab Rep.Lower middle income | 3.7%4,408 | Under 0.1%2 |
| Iran, Islamic Rep.Upper middle income | 2.7%3,214 | 0.1%8 |
| BrazilUpper middle income | 2.5%3,014 | 0.1%6 |
| ItalyHigh income | 2.5%2,981 | 1.1%61 |
| GermanyHigh income | 2.5%2,938 | 3.2%181 |
| Korea, Rep.High income | 2.1%2,545 | 2.4%135 |
| Saudi ArabiaHigh income | 1.8%2,096 | 0.2%10 |
| United KingdomHigh income | 1.7%2,052 | 3.1%177 |
| PakistanLower middle income | 1.3%1,562 | Under 0.1%1 |
| SpainHigh income | 1.2%1,406 | 0.7%38 |
| AustraliaHigh income | 1.1%1,283 | 1.9%105 |
| CanadaHigh income | 1.0%1,182 | 1.5%85 |
| PolandHigh income | 1.0%1,174 | 0.4%20 |
| NetherlandsHigh income | 0.9%1,046 | 1.0%56 |
| RomaniaHigh income | 0.9%1,014 | 0.1%5 |
| FranceHigh income | 0.8%951 | 1.1%63 |
| IndonesiaUpper middle income | 0.8%908 | 0%0 |
| Taiwan, ChinaHigh income | 0.7%887 | 0.4%24 |
| MexicoUpper middle income | 0.7%868 | 0.1%7 |
| SwitzerlandHigh income | 0.5%628 | 1.1%60 |
| GreeceHigh income | 0.5%626 | 0.1%7 |
| ThailandUpper middle income | 0.5%593 | 0.1%5 |
| MalaysiaUpper middle income | 0.5%580 | 0.1%3 |
| SwedenHigh income | 0.5%554 | 1.8%100 |
| PortugalHigh income | 0.5%543 | 0.1%3 |
| MoroccoLower middle income | 0.4%524 | 0%0 |
| NigeriaLower middle income | 0.4%515 | 0%0 |
| IraqUpper middle income | 0.4%476 | 0%0 |
| TunisiaLower middle income | 0.4%466 | 0%0 |
| BelgiumHigh income | 0.4%447 | 0.5%31 |
| DenmarkHigh income | 0.3%394 | 0.5%31 |
| NepalLower middle income | 0.3%393 | 0%0 |
| AustriaHigh income | 0.3%323 | 0.4%22 |
| ColombiaUpper middle income | 0.3%311 | Under 0.1%1 |
| IsraelHigh income | 0.3%302 | 0.3%19 |
| ArgentinaUpper middle income | 0.2%296 | 0%0 |
| FinlandHigh income | 0.2%264 | 0.3%19 |
| BangladeshLower middle income | 0.2%257 | 0.1%4 |
| Russian FederationHigh income | 0.2%256 | 0.3%17 |
| Viet NamUpper middle income | 0.2%256 | Under 0.1%1 |
| EthiopiaLow income | 0.2%251 | Under 0.1%1 |
| SingaporeHigh income | 0.2%211 | 0.6%34 |
| Hong Kong SAR, ChinaHigh income | 0.1%169 | 0.8%45 |
These figures show a concentration in published authorship. Affiliation alone cannot establish whether a laboratory has access to the technology.
The price of taking part
A spatial transcriptomics experiment can require specialised instruments, proprietary reagents, sequencing, computing and trained staff, putting it beyond the reach of many laboratories.
EPFL's Gene Expression Core Facility, for example, charges its own users about CHF 9,000 to 13,000 to run and sequence one Visium HD slide of two samples, depending on the capture area, the tissue and the sequencing depth, before histology and analysis (price list of 9 September 2026).
Another way into spatial biology
Routine pathology offers a more widely available starting point: H&E images. Hospitals and research institutions hold tissue collections that are already scanned or can be digitised.
Virtual spatial transcriptomics uses machine learning to predict spatial gene expression from these images. It offers a way to explore molecular patterns in existing tissue collections when a new assay is impractical or unaffordable. The results are predictions, and experiments remain essential to train the models and validate their findings.
In our DeepSpot-M-0 preprint, we argue that this approach could broaden biomarker discovery. Its model builds on our earlier ones: DeepSpot, which predicts expression at individual spots, and DeepSpot2Cell, which predicts it at single-cell resolution.
How we plan to act on it
We are working on ways to make this analysis available to researchers who have H&E images but lack the infrastructure for spatial assays. The aim is to help them investigate new questions in existing tissue collections and decide which findings to pursue experimentally. We will share our plans soon.
How we counted
We searched PubMed for papers published from 2021 to 2025 that mention spatial transcriptomics, assigned each to the country of the first author's first affiliation, and grouped countries using the World Bank's FY27 income classification. We compared these with H&E and histopathology papers from the same period and with random samples of PubMed records.
Affiliation is an imperfect proxy. Samples may be collected in one country and analysed in another, collaborations span institutions, and PubMed coverage varies by region.