Seven Technologies, One Lens: What 200,000 Papers Reveal About the Precision Medicine Adoption Curve
We spend a lot of time in this industry talking about where technologies are headed. Much less time measuring where they actually are. The diffusion index was built to fix that.
The premise is simple. When a technology is new and exciting, it punches above its weight in the most competitive journals in the world — Nature, Science, Cell, and the top specialist journals in its field. As it matures and diffuses, that novelty premium erodes. The ratio of top-tier papers to total papers is not just a bibliometric curiosity. It is a leading indicator of where a technology sits on the Rogers adoption curve. We have now built this index for seven precision medicine technologies, drawn from roughly 200,000 PubMed papers published between 2005 and mid-2026. Here is what we found.
The benchmark table
The cleanest way to see the pattern is to line them up. Scores are color-coded: green = novelty premium still intact, amber = maturing, red/gray = floor reached or approaching.
| Technology | Papers | Peak score | Peak year | Current score | Phase today |
|---|---|---|---|---|---|
| NGS | 112,826 | 118 / 1,000 | 2008 | ~0.8 / 1,000 | Laggards |
| Single-cell RNA-seq | 35,434 | 155 / 1,000 | 2016 | ~4 / 1,000 | Early Majority |
| Spatial transcriptomics | 6,653 | 146 / 1,000 | 2020 | ~12 / 1,000 | Early Majority |
| Spatial proteomics | 1,856 | ~32 / 1,000 | 2021–2023 | ~10 / 1,000 | Early Adopters |
| Long-read sequencing | 10,395 | 29 / 1,000 | 2017 | ~2 / 1,000 | Early Majority |
| cfDNA / Liquid biopsy † | 20,440 | ~91 / 1,000 | 2014–2017 | ~47 / 1,000 | Late Majority (clinical) |
| Multi-omics (broad) † | 37,000+ | 184 / 1,000 | 2015 | ~60 / 1,000 | Early Majority |
| High-plex multi-omics ‡ | 673 | 464 / 1,000 | 2019 | ~114 / 1,000 | Early Adopters ↑ |
† cfDNA and multi-omics (broad) scores are top-tier specialist journals (tier 1+2). Top-3 score is near-zero for both fields throughout — which is itself a finding explained below.
‡ High-plex multi-omics requires a named high-throughput method (CITE-seq, 10x Multiome, Olink+sequencing, spatial multi-omics, proteogenomics). Tier 1+2 score. 673 papers, 2017–2026. Full index →
Phase reflects the Adoption Index classification (a volume-growth state machine); the current-score column is the novelty premium, a separate axis. A field can sit in Early Majority by adoption while its top-journal score is already near the floor — single-cell and long-read are still growing in volume even though their premium has largely gone.
You can also explore all seven curves aligned to a common year zero on the normalized comparison chart.
Two classes of technology, not one
The most important finding is not the individual scores. It is that these seven technologies divide cleanly into two fundamentally different categories.
The first class — NGS, single-cell RNA-seq, spatial transcriptomics — are discovery platforms. They unlocked experiments that could not be done before. Doing NGS in 2008 was publishable in Nature not because it was expensive but because it revealed biology that was genuinely invisible before. The same is true for scRNA-seq in 2016 and spatial Tx in 2020. The adoption curve for discovery platforms follows a clean bell: high novelty premium at launch, rapid decline as the technique becomes routine, eventual floor near zero. Nature and Science reward paradigm shifts. It turns out they can tell the difference.
The second class — cfDNA, multi-omics (broad) — are clinical translation and methodology technologies. They moved existing capabilities (DNA detection, data integration) into new contexts. cfDNA never cracked 6 per 1,000 in Nature/Science/Cell despite 20,000+ papers and transforming oncology diagnostics — it got the clinical equivalent of a standing ovation in NEJM, but Nature did not particularly care. The broad multi-omics dataset peaked at 184 per 1,000 in tier 1+2 journals but is declining fast as the term becomes a genre descriptor rather than a differentiator. By 2026, appending "multi-omics" to a grant application is roughly as informative as specifying that you plan to use computers.
There is a twist, though. Strip out the loose usage — the studies that call themselves multi-omics while combining two low-plex assays — and require an actual named high-throughput method (CITE-seq, 10x Multiome, Olink + sequencing, spatial multi-omics, proteogenomics), and a completely different curve emerges. The high-plex multi-omics index covers 673 papers from 2017 to 2026, growing from 9 to 152 papers per year, with a tier 1+2 score that peaked at 464 per 1,000 in 2019 and is still at 114 today. That is not a technology in decline. That is a discovery platform in its Early Adopters phase — following the same curve shape as spatial transcriptomics, just starting a few years later. The word "multi-omics" lost its signal. The underlying science did not.
Long-read sequencing sits in an interesting middle ground: a genuine platform technology, but one that was always an extension of short-read rather than a paradigm replacement. It peaked at only 29 per 1,000 — a fraction of NGS or scRNA-seq — and has already reached its floor. The tool is powerful and widely used. It just never generated the discovery premium those others did. The infrastructure penalty is real.
The compression story
Within the discovery platform class, there is a second pattern worth noting: the adoption curve is compressing.
NGS took roughly eight years to move from Innovators to the floor (2008 → 2016). Single-cell RNA-seq took about ten years and is only now approaching the floor. Spatial transcriptomics peaked in 2020 and is already at 12 per 1,000 six years later — on track to reach its floor faster than either predecessor.
Part of this is the platform effect: each new generation inherits the infrastructure, vocabulary, and institutional knowledge of the previous one. Bioinformaticians who analyzed NGS data learned scRNA-seq faster. Labs equipped for scRNA-seq adopted spatial more quickly. The learning curve compounds in reverse.
But part of it is also market maturity. The life sciences ecosystem now has better tooling, better commercial partners, and better clinical-translational infrastructure for absorbing new measurement technologies. What took a decade in 2008 takes five to six years now. The implication for investors and strategists: the window to capture the premium phase of a new precision medicine technology is shorter than it used to be. It is also shorter than most five-year strategic plans assume.
What the scores do not capture
The diffusion index measures scientific novelty. It does not measure commercial value or clinical impact. These are different things, and conflating them is a mistake.
cfDNA's flat top-3 score is not an indictment of the technology. It is an indictment of a category assumption: that the way to assess a clinical platform's value is through the same lens you would use for a discovery technology. Guardant360, FoundationOne Liquid CDx, and Signatera collectively run millions of tests per year. The score just tells you that the scientific community is no longer surprised by the technique. That is a feature, not a bug — it means the technology has successfully transitioned from laboratory curiosity to clinical workflow.
Similarly, the collapse in multi-omics score does not mean multi-omics studies are less important. It means the field has internalized the approach. A multi-omics paper in 2015 was doing something genuinely novel. A multi-omics paper in 2026 is describing a standard methodological choice, like saying "we used PCR." The word lost signal because the practice became universal.
Where this matters for strategy
For anyone positioning a company, allocating research funding, or evaluating a market:
The discovery platform with a high score and a rising curve is the one worth investing in early. Spatial proteomics — IMC, MIBI, CODEX, CyCIF, GeoMx — is at 10 per 1,000 and still in Early Adopters. That is roughly where spatial transcriptomics was in 2022. The market has seen what happened next with spatial Tx.
The platform approaching its floor (long-read sequencing at ~2/1,000, scRNA-seq at ~4/1,000) is one where the moat shifts from scientific novelty to workflow integration, cost, and scale. The companies that win at floor are not the ones generating the most exciting papers. They are the ones building the most reliable, integrated, and economically defensible infrastructure.
And the clinical translation technology with a stable, non-zero floor (cfDNA at ~47/1,000 in tier 1+2) is one that keeps generating evidence for new indications — MRD in solid tumors, MCED, transplant monitoring — without the novelty premium resetting. That evidence generation is what supports ongoing reimbursement coverage and clinical adoption. The diffusion score stabilizing at 40–50 per 1,000 is not a technology in decline. It is a technology that has become operational.
The index as a leading indicator
We built this index because the usual proxies for technology maturity — VC funding, conference buzz, analyst coverage — are all lagging. By the time a technology is on the cover of Nature Reviews or the subject of a STAT News special report, the Innovators phase is over and the Early Majority is already arriving.
Papers are leading. The ratio of top-tier papers to total papers is a real-time measurement of how much the scientific community is still treating a technique as novel. When that ratio peaks, the chasm is approaching. When it collapses, the mainstream has crossed. The seven curves here are not predictions. They are measurements — of something the market is always a few years behind in pricing.
The full interactive diffusion index for all seven technologies is on the Signal page. The normalized year-zero comparison chart is here, and the Adoption Index places all eight technologies on the Rogers curve directly — including how the framework was calibrated against what we know to be true in the field.