The Diffusion Index
The NGS Diffusion Atlas
From Methods to Medicine
Next-generation sequencing did not just grow — it spread. It began as a methods-and-bioinformatics frontier and worked its way, application by application, into oncology, prenatal testing, liquid biopsy, and the rest of clinical medicine. This view traces that spread across 98,417 classified NGS publications, 2006–2025.
8% → 49%
Clinical & diagnostic share of NGS papers (2008 → 2025)
2.9 → 7.1
Effective applications — the breadth NGS is spread across (2008 → 2025)
10 / 11
Applications at mainstream scale (100+ papers in 2025)
98,417
NGS papers classified by application, 2006–2025
A scholarly diffusion view: it measures how NGS spread through the published research community, classified by application from each paper’s title and abstract. It is not a measure of clinical or commercial adoption. Data through June 2026; 2026 is partial and excluded from this view.
When each application lit up
A dot each time an application’s cumulative output crossed an order of magnitude — 1, 10, 100, 1,000, 10,000 papers — with a final dot at the present. Tightly-spaced rings mean explosive growth; stretched-out rings mean a slow climb. Lanes run from basic research at the top, through translational and clinical, down to applied markets — read top to bottom and the rings drift rightward: research arrived first, the clinical and diagnostic applications followed, most within a decade. Toggle the basis between cumulative totals and annual output and the rings reposition — the year a field’s running total passed 1,000 is not the year it first published 1,000 papers in a single year.
Each dot = the year cumulative output crossed an order of magnitude; the final dot is the all-time total.
From research into the clinic
The clearest single signal of NGS becoming infrastructure: the share of NGS publications doing clinical and diagnostic work climbed from 8% in 2008 to 49% by 2025. Meanwhile the number of effective applications — how many distinct uses the literature is meaningfully spread across — more than doubled, from 2.9 to 7.1.
“Clinical & diagnostic” combines clinical Dx (oncology, NIPT, other), liquid-biopsy diagnostics, and biopharma/liquid-biopsy clinical trials. “Effective applications” is the exponential of the Shannon entropy of the application mix (1 = a single use; 11 = perfectly even across all categories).
How fast each application diffused
Years from an application’s first NGS paper to its first year at mainstream scale. A short gap means the use case was waiting for the technology; a long one means the science (or the clinical evidence) had to catch up.
| Application | First paper | Reached 100/yr | Years to mainstream | 2025 |
|---|---|---|---|---|
Academia / basic research | 2006 | 2009 | 3 | 1,372 |
Population studies | 2006 | 2011 | 5 | 315 |
Other applications | 2006 | 2011 | 5 | 2,336 |
Clinical Dx — other | 2006 | 2012 | 6 | 2,688 |
Biopharma R&D | 2007 | 2014 | 7 | 535 |
Clinical Dx — oncology | 2006 | 2013 | 7 | 1,734 |
Epidemiology / infectious dz | 2006 | 2014 | 8 | 838 |
Biopharma clinical trials | 2012 | 2021 | 9 | 136 |
Liquid biopsy — clinical trials | 2013 | 2022 | 9 | 95 |
Liquid biopsy — diagnostics | 2009 | 2019 | 10 | 273 |
Clinical Dx — NIPT / prenatal | 2009 | 2019 | 10 | 168 |
From publications to dollars
The Atlas reads scholarly activity. Does it track the money? We also model the NGS market bottom-up — and the two instruments agree. Expressed in the same unit, the share of NGS that is clinical, publications (49% in 2025) and manufacturer revenue (~51%) have converged on roughly half. The publication curve is the 15-year climb that got there first.
Publication share from the Atlas. Revenue share from DeciBio’s NGS Market Report (11th ed., Dec 2025; figures as shared publicly in the Cowen investor presentation), which sizes the NGS manufacturer market — instruments, reagents, bioinformatics — at ~$7.3B in 2025 (clinical ≈ $3.7B) and excludes diagnostic service revenue. 2028 is projected from the report’s published clinical growth rates. Two different things (research activity vs. tool revenue), one shared axis.
Atlas signal
Clinical & diagnostic share of NGS papers rose from ~8% (2008) to 49% (2025).
Market confirms
Clinical is now ≈$3.7B of the ~$7.3B market and ~80% of all market growth to 2028.
Atlas signal
Liquid-biopsy applications emerged late (first papers 2009–2013) and are still climbing.
Market confirms
MRD / liquid biopsy is the fastest-growing demand — Natera Signatera grew 54% YoY to ~211k tests in Q3 2025.
Atlas signal
The research lanes (academia, population) are the oldest and now the slowest-growing.
Market confirms
Academia is ~25% of revenue but growing only ~1%/yr — research publishes, the clinic pays.
Atlas signal
Effective applications more than doubled (2.9 → 7.1) — NGS spread across many uses.
Market confirms
Multiple engines now drive demand — rare-disease WGS/WES (GeneDx >30% growth) and women’s health beyond NIPT.
And the value accrued to diagnostics, not tools: through Dec 2025, Guardant Health was up 221% and Natera up 52% on the year while Illumina was down 2% — even as sequencing data volumes grew 30–40%/yr and reagent revenue rose only ~3%/yr. The reads got cheaper; the clinical decisions they enable got more valuable. (DeciBio NGS Market Report / Cowen, Dec 2025.)
What this measures (and what it doesn’t)
Every NGS paper with an abstract is assigned an application area by a language model (Claude Haiku, via the Batch API) from its title and abstract. Because the classification reads the science itself — not author affiliations — this view carries none of the first-author or institution-resolution bias that an institution-level analysis would.
It is a measure of scholarly diffusion: where the published research community is putting NGS to work. It is a leading indicator, not a measure of clinical or commercial adoption — a field can be heavily published before it is widely reimbursed, and vice-versa. “Mainstream” here is a deliberately simple, transparent bar (100+ papers in a single year), chosen so the milestones are auditable rather than tuned. Papers that span several applications are assigned a primary area, so counts are conservative for genuinely cross-cutting work.
Deliberately out of scope (for now): institution- and author-level diffusion — who entered the field, collaboration networks, and bridge institutions — which the current dataset (first-author affiliation strings, no coauthorship) can’t support credibly. That would need a structured authorship source such as OpenAlex.