The Diffusion Index
cfDNA & Liquid Biopsy
20,440 PubMed papers. The diffusion curve for cell-free DNA looks nothing like NGS or single-cell — and that gap is the most revealing finding. Clinical translation technologies follow a different pattern than discovery platforms.
Papers analyzed
20,440
Peak tier 1+2 (2007)
1000/1,000
Top-3 peak (ever)
~6/1,000
Current (2026)
47/1,000
How many papers, and where they land
Annual cfDNA 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. For a clinical-translation field, the top-3 line barely lifts off the floor — which is itself the story the diffusion curve below tells 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).
Peak: 500 per 1,000 in 2007 (blended) · current (2026): 24.4 per 1,000 — settling into routine clinical use
The flat top-3 signal is the finding
In 15 years of cfDNA research, the top-3 journal score has never exceeded ~6 per 1,000. For comparison, NGS peaked at 118/1,000, single-cell at 155/1,000, and spatial at 146/1,000. The gap reflects a fundamental difference: those technologies unlocked new experiments that could answer questions biology hadn’t previously been able to ask. cfDNA did something different — it moved an existing assay (DNA detection) from tissue to blood. Clinically transformative, yes. But not a paradigm shift in how we do science. Nature and Science reward paradigm shifts.
Clinical journals tell the real story
The tier 1+2 score — which includes NEJM, Lancet, JCO, and other clinical heavyweights — peaked at ~91 per 1,000 around 2014 and 2017. These peaks mark the clinical validation era: the pivotal studies that proved cfDNA could detect cancer, guide treatment, and monitor recurrence. Guardant360 launched commercially in 2016; FoundationOne Liquid CDx followed. From there, the novelty premium declined steadily as cfDNA became a standard clinical tool rather than a research frontier.
ddPCR to NGS: two eras
Platform detection in the underlying papers reveals two distinct technological eras. Digital droplet PCR (ddPCR, 1,072 mentions) dominated early cfDNA research — it was sensitive enough to detect rare mutations but single-target. NIPT (776 mentions) shows how prenatal cfDNA drove massive adoption from the clinical side before oncology did. The sequencing era (Guardant 169, Natera/Signatera 76) came later but at scale: these commercial platforms now collectively run millions of tests annually, most of which never appear in a paper.
The research-to-clinical divergence
Since 2022, cfDNA paper volume has plateaued around 2,000–3,000 per year while commercial test volumes have continued growing. The diffusion score (~47/1,000 in 2026) has stabilized rather than continuing to fall — unlike discovery technologies which hit near-zero floors. This suggests cfDNA has reached a permanent equilibrium: ongoing clinical evidence generation in NEJM, JCO, and Lancet for new indications (MCED, MRD, transplant), but no longer generating fundamental science surprises. The technology is operational. The papers are incremental.
Discovery technologies vs. clinical translation
The diffusion index distinguishes two fundamentally different technology categories. The curve shape tells you which kind you’re looking at.
Discovery platforms (e.g. NGS, scRNA-seq)
Top-3 peak: 100–160/1,000
Classic bell curve with steep initial rise
Floor: 0.8–4/1,000 (near zero)
Nature/Science/Cell reward the method itself
Clinical translation (e.g. cfDNA, ctDNA)
Top-3 peak: ~6/1,000 (essentially flat)
Tier 1+2 shows a real but muted curve
Floor: ~47/1,000 (clinical journals ongoing)
Evidence generation never stops — it just becomes routine
What the sequencing was for
The application mix of cfDNA papers, 2008–2025, classified into research and clinical uses and normalized to 100% per year. cfDNA is overwhelmingly a clinical field — and the clinical frontier keeps moving.
Research
Clinical
Normalized to 100% per year. Classified by Claude (Haiku) over 19,208 cfDNA papers into tech-specific buckets. 2008–2025.
cfDNA began as a prenatal-testing and assay-biology field: through about 2014, non-invasive prenatal testing (NIPT) and methods work were the bulk of output. Oncology therapy selection — genotyping ctDNA to choose a treatment — overtook NIPT around 2015–2016 and is now the single largest use (~27% in 2025).
The real signal is the frontier. Minimal residual disease (MRD) monitoring and multi-cancer early detection (MCED) were near-zero before 2017; together they are now roughly a third of cfDNA papers and the fastest-growing slices, while NIPT — the original killer app — has shrunk to ~7%. This is exactly why a single “phase” for cfDNA misleads: it is mature in therapy selection, still climbing in MRD, and early in MCED — all at once.
Who uses it for what
cfDNA’s application mix crossed with first-author geography. Click regions to set the comparison baseline and see who over-indexes on which use — therapy selection, MRD, early detection, or prenatal testing.
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 | ||||||
| Methods & cfDNA biology | 16.1% | 14.1% -12% | 15.3% -5% | 11.0% -32% | 19.6% +22% | 19.1% |
| Population / epidemiology | 1.2% | 1.3% +10% | 1.3% +8% | 1.2% -4% | 1.0% -18% | 1.1% |
| Clinical | ||||||
| Oncology — therapy selection | 28.8% | 28.8% +0% | 30.5% +6% | 40.8% +42% | 23.8% -17% | 24.1% |
| Oncology — MRD / recurrence | 15.6% | 14.2% -9% | 17.0% +9% | 21.5% +38% | 13.5% -14% | 12.2% |
| Oncology — early detection / MCED | 14.4% | 12.1% -16% | 12.7% -12% | 11.6% -19% | 19.1% +33% | 12.8% |
| NIPT / reproductive | 9.3% | 11.4% +22% | 9.7% +4% | 5.0% -46% | 8.1% -14% | 8.7% |
| Transplant monitoring | 3.5% | 6.8% +92% | 3.6% +1% | 1.2% -67% | 1.4% -59% | 3.9% |
| Infectious disease / sepsis | 4.4% | 7.1% +61% | 3.1% -29% | 3.0% -33% | 4.4% +1% | 4.0% |
| Other clinical | 6.7% | 4.3% -37% | 6.8% +0% | 4.8% -29% | 9.0% +34% | 14.0% |
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 18,945 classified papers with a parseable first-author affiliation; region assigned by keyword match.
Where the papers come from
Share of cfDNA papers by first-author affiliation, 2008–2026. Parsed from 20,060 affiliations.
cfDNA is the geographic outlier. No single country has ever exceeded ~30% of first-author output, and rest-of-world has been the majority every year since 2008 — France, the Netherlands, Spain, and others contributed heavily to early noninvasive-prenatal and liquid-biopsy work. China has run ahead of the US for most of cfDNA’s history, reflecting its early and large investment in NIPT. This is a clinically driven, globally distributed field — not a two-nation race.
Methodology: Papers fetched from PubMed matching “cell-free DNA” OR “cfDNA” OR “circulating tumor DNA” OR “ctDNA” OR “circulating free DNA” OR “plasma DNA sequencing”. Excludes liquid biopsy papers focused on CTCs or exosomes. Default view shows Tier 1+2 journals (NEJM, Lancet, JCO, Nature Medicine, etc.) — the more informative metric for a clinical-translation technology. Journal tiers assigned locally using a curated list. Data as of June 2026.