Digital PCR value assignment at 0.001%-0.01% VAF — Poisson floor, input mass, replicate counts and published uncertainty budgets

1. Summary

The mass-to-copy conversion used in the metrology literature is 3.3 pg per human haploid genome, which is roughly 300 genome equivalents per ng [1][2]. The NMI dPCR studies read here report limits of quantification or detection at or above about 0.1% VAF (variant allele frequency), and none reports a value assignment below that level [6][7][9][10][14]. No NMI or vendor document was found that value-assigns a ctDNA material at 0.001%-0.01% VAF with a stated expanded uncertainty; the lowest found, and its single source, are in section 3.

2. What Changed

_Baseline (v1). No prior version; this establishes the starting point for future diffs._

3. Details

Conversion constant and sampling at low VAF

• LGC (2014): "a single human haploid genome that is calculated as 3.3 pg" [1] [single-source].
• The FNIH ctDNA quality-control study (JCO Precision Oncology, 2021) used materials at 5%, 2.5%, 1%, 0.1% and 0% VAF. Two ddPCR sites took part: NIST (9 ng input, "approximately 2,700 genomic copies") and Dana-Farber Belfer (16 ng) [2] [single-source].
• The same study states: "A VAF of 0.1% has about 12 variant molecules in a 40 ng input." It also states: "Dropouts occurred in all assays across all variant types at the 0.1% VAF level." The authors add that this is "reasonable … (< 3 positives on average expected in 9 ng and five positives in 16 ng)" [2] [single-source].

What NMIs have value-assigned, and how low

• KRAS G12D dPCR was assessed as a primary reference measurement procedure (Clin Chem 2018). Repeatability CV was 2%-8% and interlaboratory CV 5%-10%. The paper assigned "copy number concentration and fractional abundance values to DNA reference materials in an aqueous solution" [5] [single-source]. This was read at abstract depth, and the abstract gives no VAF floor.
• BRAF V600E is an SI-traceable, gravimetric "ctDNA" reference material (Talanta 2020). ddPCR agreed with the gravimetric value (k = 1.03) "in the range from 53.9% to 0.1%", with LoB/LoD/LoQ of 0.01%/0.02%/0.1% [6] [single-source].
• EGFR L858R/T790M/19Del dPCR primary reference measurement procedure (Anal Methods 2021): linear from 25% to 0.05% FA. The LoQ was "34 L858R, 23 T790M, and 34 19Del copies/reaction, corresponding to a FA of 0.2%" [7] [single-source].
• KRAS NIM reference material (Sci Rep 2018): the ddPCR LoQ was 0.1% [14] [single-source].
• NIM, KRISS and NMIJ co-validated KRAS G12A fractional abundance by dPCR. All En values were ≤1 "within the expanded uncertainty" [8] [single-source]. The FA level and the size of the uncertainty are not given in the abstract.
• KRISS nucleosomal EGFR reference materials (Anal Bioanal Chem, published 2026-04-06, open full text). The lowest assigned level (Level 2) was T790M 0.27% (U 0.09, 33%), L858R 0.25% (U 0.15, 59%), G719S 0.32% (U 0.11, 35%) and Ex19del 0.26% (U 0.24, 90%). The uncertainty components were characterization, between-bottle heterogeneity and manual thresholding. Level 1 was reported as "not detected" [9] [single-source].
• NRAS Q61R candidate reference measurement procedure (Methods, 2026): LoD 0.1%, with "a full measurement uncertainty budget" [10] [single-source]. The budget itself was not read.
• NIST's multi-laboratory study of a commercial 40-variant material (JMD 2019) used dPCR to measure VAFs at 2%, 0.5%, 0.125% and 0% [13] [single-source].
• NIST SRM 2373 is a HER2 genomic-DNA copy-number-ratio standard (certified ratios 1.3 to 17.7), not a ctDNA VAF material [3] [single-source]. NIST's completed ctDNA project page says only that required allele frequencies are "as low as 0.01%". It names no released material number [4] [single-source].

Vendor material at MRD levels

• The Seraseq ctDNA MRD Panel Mix (catalog 0710-2146) has tumor blends of 0.5%, 0.05% and 0.005%. dPCR was run only on "0.5% tumor content equivalent = 0.25% VAF", pre-fragmentation, with a six-variant average of 0.43% VAF. At 0.005% tumor, NGS (next-generation sequencing) observed 4 variant copies across 4 of 100 variants. No uncertainty is stated at any level [12] [single-source · manufacturer].

Partitions and false positives

• The KRISS 2026 study used more than 10,000 accepted droplets per QX200 reaction and 10 replicates over 3 days. λ for wild-type was 0.12-0.17. LoB, taken from more than 85 wild-type reactions, was 0.04%-0.11% VAF, and LoD was below 0.15% [9] [single-source].
• The EGFR procedure reported a "false positive rate 0-0.003%" [7] [single-source]. The BRAF LoB was 0.01% [6] [single-source].
• Across 6 Bio-Rad ddPCR assays, polymerase-induced false-positive events were found. Correcting them together with an adaptive LoB raised specificity in healthy cfDNA (cell-free DNA) from 88% to 98% [11] [single-source].

4. Sources

[1] Towards standardisation of cell-free DNA measurement in plasma: controls for extraction efficiency, fragment size bias and quantification — Analytical and Bioanalytical Chemistry (published 2014, day not retrieved; accessed 2026-09-25). PMID (PubMed identifier) 24853859; DOI 10.1007/s00216-014-7835-3 — https://pmc.ncbi.nlm.nih.gov/articles/PMC4182654/ [peer-reviewed] [2] Validation of ctDNA Quality Control Materials Through a Precompetitive Collaboration of the Foundation for the National Institutes of Health — JCO Precision Oncology (published 2021-06-01; accessed 2026-09-25). PMID 34250423; DOI 10.1200/PO.20.00528 — https://pmc.ncbi.nlm.nih.gov/articles/PMC8232894/ [peer-reviewed] [3] NIST HER2 Genomic DNA Standard Reference Material SRM 2373 — NIST (published 2016-10-27, updated 2026-05-20; accessed 2026-09-25). catalog no. SRM 2373 — https://www.nist.gov/programs-projects/nist-her2-genomic-dna-standard-reference-material-srmr-2373 [other] [4] Measurement of low frequency mutations in circulating tumor DNA (ctDNA) — NIST project page (published 2016-07-01, updated 2025-03-26; accessed 2026-09-25). [no-identifier] a program web page that no registrar numbers and that names no material — https://nist.gov/programs-projects/measurement-low-frequency-mutations-circulating-tumor-dna-ctdna-through [other] [5] Assessment of Digital PCR as a Primary Reference Measurement Procedure to Support Advances in Precision Medicine — Clinical Chemistry (published 2018-09, day not given in PubMed record; accessed 2026-09-25). PMID 29903874 — https://pubmed.ncbi.nlm.nih.gov/29903874/ [peer-reviewed] [6] Interlaboratory assessment of droplet digital PCR for quantification of BRAF V600E mutation using a novel DNA reference material — Talanta (published 2020-01-15; accessed 2026-09-25). PMID 31594564 — https://pubmed.ncbi.nlm.nih.gov/31594564/ [peer-reviewed] [7] Establishment of primary reference measurement procedures and reference materials for EGFR variant detection in non-small cell lung cancer — Analytical Methods (published 2021-05-13; accessed 2026-09-25). PMID 33870958 — https://pubmed.ncbi.nlm.nih.gov/33870958/ [peer-reviewed] [8] International co-validation on absolute quantification of single nucleotide variants of KRAS by digital PCR — Analytical and Bioanalytical Chemistry (published 2022-08, day not given in PubMed record; accessed 2026-09-25). PMID 35810409 — https://pubmed.ncbi.nlm.nih.gov/35810409/ [peer-reviewed] [9] Development and characterization of nucleosomal DNA-based reference materials for the epidermal growth factor receptor gene liquid biopsy — Analytical and Bioanalytical Chemistry (published 2026-04-06; accessed 2026-09-25). PMID 41936644; DOI 10.1007/s00216-026-06475-5 — https://pmc.ncbi.nlm.nih.gov/articles/PMC13221357/ [peer-reviewed] [10] New candidate reference measurement procedure for NRAS p.Q61R quantification by digital PCR — Methods (published 2026-10, day not given in PubMed record; accessed 2026-09-25). PMID 42431334 — https://pubmed.ncbi.nlm.nih.gov/42431334/ [peer-reviewed] [11] An Automated Correction Algorithm (ALPACA) for ddPCR Data Using Adaptive Limit of Blank and Correction of False Positive Events Improves Specificity of Mutation Detection — Clinical Chemistry (published 2021, month not retrieved; accessed 2026-09-25). PMID 33842952 — https://pubmed.ncbi.nlm.nih.gov/33842952/ [peer-reviewed] [12] Seraseq ctDNA MRD Panel Mix Datasheet MKT-00626-01 — SeraCare Life Sciences (published 2021, day not given on the document; accessed 2026-09-25). catalog no. 0710-2146 — https://www.seracare.com/globalassets/seracare-resources/ds-mkt-00626-0710-2146-seraseq-ctdna-mrd-panel-mix.pdf [manufacturer] [13] Multilaboratory Assessment of a New Reference Material for Quality Assurance of Cell-Free Tumor DNA Measurements — The Journal of Molecular Diagnostics (published 2019-07, day not given in PubMed record; accessed 2026-09-25). PMID 31055023 — https://pubmed.ncbi.nlm.nih.gov/31055023/ [peer-reviewed] [14] Evaluation of droplet digital PCR and next generation sequencing for characterizing DNA reference material for KRAS mutation detection — Scientific Reports (published 2018-11-30; accessed 2026-09-25). PMID 30504843 — https://pubmed.ncbi.nlm.nih.gov/30504843/ [peer-reviewed]

5. Sourcing & Gaps

Well established: Two sources agree on about 300 GE/ng: 3.3 pg per haploid genome [1], and 9 ng ≈ 2,700 copies [2]. The NMI dPCR quantitation and detection floors read here sit at or above about 0.1% VAF [6][7][9][10][14].

Thin: Every item in section 3 rests on one paper and carries [single-source]. The vendor datasheet is [manufacturer] [12]. Sources [5], [6], [7], [8], [10], [13] and [14] were read at abstract depth only. Uncertainty budgets, FA levels and partition counts in their full texts were not read.

Rescoped from class 3: None in this report. The question is entirely class 1/2 arithmetic and published metrology.

Out of scope: Paywalled full texts of [5], [6], [8] and [10]. The NRAS budget [10] is the one most likely to hold a sub-0.1% uncertainty statement. Also out of scope: NIM China GBW certificates, JRC ERM certificates, and the certificates of analysis for commercial materials (Seraseq, Horizon/Revvity, Twist).

Not searched vs. not found: Not found: any PTB or JRC ctDNA VAF value-assignment paper (PubMed queries plus one JRC web search), and any NMI or vendor value assignment below 0.1% VAF with a stated expanded uncertainty. Not searched: Horizon/Revvity and Twist datasheets, BIPM CCQM key-comparison reports (for example CCQM nucleic-acid comparisons), and non-Bio-Rad platform partition counts. "PTB" and "LGC NML ctDNA" were not searched separately by name in PubMed. Institution attribution for [5], [6], [7] and [14] was not retrieved and is not stated.

[inference] Worked arithmetic, using 1,000 pg ÷ 3.3 pg = 303 GE/ng [1]. Mutant copies N = GE × VAF. The Poisson relative standard uncertainty is 1/√N, and P(0 copies) = e^−N.

[inference] A 10% Poisson-only relative standard uncertainty (N = 100) needs about 3.3 µg at 0.01% VAF and about 33 µg at 0.001%. This is before adding LoB subtraction, partition-volume, thresholding or heterogeneity terms. At the 0.04%-0.11% LoB reported for QX200 [9], or even at a 0.003% false-positive rate [7], background at 0.001% would equal or exceed signal unless it is characterized and subtracted.

[inference] At the KRISS loading (λ_WT 0.12-0.17 over at least 10,000 droplets [9]), one well holds about 1,200-1,700 wild-type copies, roughly 4-6 ng. Reaching 3.3 µg would then take about 600-800 wells. At λ ≈ 1 with 20,000 droplets (an assumed loading, not sourced), it would take about 50 wells.

6. Claim Candidates