This article exists because a reader-shared PDF sent me down a rabbit hole. What follows separates what is independently verifiable from what rests on a single, non-peer-reviewed reanalysis — and names the plausible non-institutional explanation (sequencing cross-contamination) alongside the institutional one, rather than picking whichever fits the narrative.
The Genomic Contamination Question
A letter circulating under the title "The bloodstream of mRNA vaccinated individuals shows DNA expression vector contamination" claims that bacterial plasmid DNA — the manufacturing scaffold used to produce spike mRNA, including SV40-associated sequences and a kanamycin-resistance gene — turns up in the blood of vaccinated people, not just in the vials. It is not a new study. It is a reanalysis of publicly deposited sequencing data from four legitimate, peer-reviewed papers that were investigating something else entirely. Here is what that means, and what it doesn't.
What the Letter Claims
The author, a bioinformatician working alone and publishing without a journal or DOI, downloaded raw sequencing reads that four other research teams had deposited in public archives (SRA — Sequence Read Archive) as a condition of their own publication. He reprocessed those reads using standard genome assembly software (SPAdes) looking specifically for fragments of the DNA expression vector — the bacterial plasmid backbone that mRNA vaccine manufacturers use to produce the spike mRNA in bulk. That vector's presence in the vials themselves, in nanogram-to-microgram quantities, had already been reported by McKernan et al. in 20231 — a finding that has drawn regulatory attention independent of this letter and sits on firmer ground than anything that follows. That preprint has since cleared peer review: Speicher, Rose and McKernan published a quantitative follow-up in Autoimmunity in September 20256, testing 32 vials across 16 lots from Ontario pharmacies with two independent methods (Qubit fluorometry and qPCR), plus Oxford Nanopore sequencing to size the DNA fragments (mean 214bp, up to 3.5kb) and preliminary DNase I sensitivity testing to check whether the DNA is protected inside the lipid nanoparticles. The SV40 promoter-enhancer-ori sequence — detected only in Pfizer vials, not Moderna — exceeded the regulatory limit by qPCR in 2 of 6 Pfizer lots tested. This is no longer a preprint claim. It is published, peer-reviewed, and quantified.
What's new in this letter is the claim that the same vector fragments — SV40 regulatory sequences, kanamycin-resistance genes — show up not just in the vials, but in blood drawn from vaccinated people, sometimes as late as 14 days after injection.
The Four Source Studies — What They Actually Looked For
This is the part worth sitting with. None of the four studies whose data got reused were investigating DNA contamination. Their raw reads were reused for a question they were never designed to answer.
Odak et al. (2024) — Germany
A systems-biology study asking why some people mount a strong immune response to the BNT162b vaccine and others a weak one. High- and low-responder women were profiled with flow cytometry, bulk and single-cell RNA sequencing, and cytokine panels. The paper's conclusions are entirely about early molecular signatures that predict later antibody and T-cell response — nothing about vector DNA.
Ryan et al. (2022 preprint → 2023) — Australia
A 102-person systems-immunology comparison of mRNA (BNT162b2/mRNA-1273) versus adenoviral-vector (ChAdOx1) vaccines, recruited during a period of zero community COVID-19 transmission in South Australia — meaning any immune signature seen could be attributed to the vaccine, not incidental infection. The finding that made headlines from this paper: ChAdOx1, not the mRNA vaccines, produced a memory-like innate response tied to complement and coagulation proteins. Again, nothing about DNA vector contamination — that question was never asked.
Lee et al. (2022) — South Korea
A 46-person cohort comparing mixed ChAdOx1-BNT162b2 vaccination against two doses of BNT162b2 alone. Found the mixed schedule produced a stronger antibody and T-cell response, including against Omicron. A vaccine-effectiveness comparison, not a contamination study.
Knabl et al. (2022) — Austria
A different design entirely: hospitalized COVID-19 patients infected with the Beta variant, comparing four who'd had one BNT162b dose against five unvaccinated. RNA-seq showed vaccinated patients mounted a stronger interferon/JAK-STAT antiviral response at day 10 — a study about vaccination's effect during active infection, not about vector persistence in healthy vaccinees.
Two Explanations, Deliberately Left Open
To his credit, the letter's author does not claim to have solved this. He names two competing explanations for the vector DNA reads, side by side, without picking one:
- Manufacturing contamination at the mRNA level — the vector DNA was converted into RNA during production and is present as residual contamination in the finished product, circulating and clearing like any other foreign RNA.
- Intact DNA integration — the vector DNA remains as DNA, enters the nucleus of cells, and potentially inserts into the genome. This is the explanation that would matter for long-term risk. It is also the one with zero direct evidence behind it in this letter or anywhere else.
There is a third possibility the letter doesn't raise, and it deserves equal weight: cross-sample contamination during sequencing itself. When labs run many samples through the same sequencer in the same batch — standard practice, not a shortcut — a small percentage of reads from one sample routinely bleed into another's data file. This is a known, studied artifact in metagenomic and viral-reads-in-blood research generally, and it would produce exactly the pattern seen here: low-level vector reads appearing where they shouldn't, at levels too low to represent a biologically meaningful dose. This explanation requires no institutional motive on anyone's part — it's a housekeeping problem in shared sequencing infrastructure. It doesn't rule out the other two. It just means "we found reads" is not the same claim as "this is in your blood at a dose that matters," and none of the three explanations has been ruled out by anyone yet.
UPDATE (10 August 2026) — McKernan Pushes Back, With Receipts From His Own Prior Work
Kevin McKernan — whose own 2023 finding of plasmid DNA in the vaccine vials opens this article — replied directly to a post about this piece. Two things from that reply, plus a correction of our own, change the picture above.
First, he's already taken this to regulators. He states he has disclosed these same five papers (Lee, Knabl, Odak, Ryan, Chakraborty) to ACIP, the US CDC's own vaccine advisory committee, as evidence of contamination. This isn't a single unreviewed voice shouting into the void — it has already been formally raised with the people who set vaccine policy.
Second, the cross-sample contamination explanation above doesn't survive his technical objection. McKernan points out that many of the reads bridge vector and spike sequence — meaning a single continuous fragment contains both — a pattern that simple index-hopping between samples doesn't produce. Cross-contamination creates isolated foreign reads landing where they shouldn't; it doesn't stitch two different sequences from two different samples into one chimeric fragment.
Correction: we originally attributed a more detailed technical rebuttal to "an independent analysis" published under the name Anandamide on Substack ("Chakraborty Part II: Odak et al."). That attribution was wrong — Anandamide is McKernan's own newsletter (kevinmckernan.substack.com). He had already worked through this same dataset in detail back in December 2024, months before this exchange, and reached the same conclusion there: the DNBSEQ G400 sequencer used in this data circularizes DNA into nanoballs before amplification, a method that suppresses index-hopping by 100-1000x compared to standard platforms — ruling out simple cross-sample contamination on hardware grounds. His conclusion in that piece: the more likely explanation is incomplete DNA depletion during library preparation — the vector DNA simply wasn't fully removed before sequencing. This is one source making the same case twice, in more depth the first time, not two independent parties arriving at it separately — worth being precise about, even though it doesn't weaken the technical point itself.
McKernan does still agree with one specific caveat already flagged in Chakraborty's own letter: the SV40 signal found in the Moderna vials specifically is likely an artifact, given it was sequenced on the same machine as Pfizer samples at low reads-per-million.
What this changes: the "just a sequencing artifact" explanation offered above was too confident. The presence of vector DNA reads in blood samples now looks more likely to reflect something real — either residual contamination in the finished product itself (consistent with McKernan's own vial-level work) or incomplete purification during lab processing — rather than a batch-contamination quirk.
What hasn't changed: genomic integration is still unproven. Neither Chakraborty nor McKernan has demonstrated that vector DNA has inserted into a human genome. That the DNA is present looks more solid than it did before this update. That it integrated remains an open hypothesis — the tier-open dot in the legend above still applies to that specific claim, and only that claim.
UPDATE 2 (9 August 2026) — McKernan Explains the Signal Loss, Real Gene-Expression Evidence Surfaces, and an Independent Lab Confirms Detection
The exchange kept going. Three things from it are worth recording here rather than leaving scattered across X.
McKernan's mechanism for why the signal is patchy across studies: he clarified he wasn't dismissing incomplete DNA depletion as an explanation, just noting it doesn't account for everything. His more specific point: the globin-depletion step that labs routinely use to strip overwhelming hemoglobin mRNA out of whole-blood RNA-seq has sequence homology to the BNT162b2 5'UTR — because Pfizer and Moderna both built their vaccine UTRs from globin gene sequences in the first place (a design detail Chakraborty's own letter documents). That means the standard purification step used across these studies likely strips out some of the real vaccine signal along with the hemoglobin mRNA. If true, the actual level of vector material in blood is being undercounted by this methodology, not inflated by it. He also pointed to two further arguments: a dose-response pattern (more reads in patients sequenced shortly after vaccination) visible across the Odak cohort's own timepoints, and the presence of pre-vaccination baseline samples as an internal negative control — if index-hopping or reagent-kit contamination explained the signal, it should show up in the "pre" samples too, and in the underlying tables it largely doesn't.
A separate reanalysis McKernan linked (his own, of an Austrian elderly-cohort study) reports something more directly useful than any of this DNA-detection back-and-forth: measured gene-expression evidence. Five genes downstream of the cGAS-STING pathway — the same DNA-sensing alarm system this article's Genomic Sentinel Protocol targets — were found elevated seven days after vaccination in that cohort. That moves the cGAS-STING mechanism in this piece from "biologically plausible, independent of cause" to "measured, in at least one reanalyzed dataset." It doesn't touch the integration question. It does support the practical, protocol-level relevance of addressing that alarm response regardless of how the underlying DNA question resolves.
Independent confirmation, from a lab, not a reanalysis: Dr. Rogier Louwen — a virologist who has previously co-authored forensic sequencing work with this site5 — reported testing 200 of his own samples and finding SV40 sequence present in 10-15% of them, "confirmed with sequencing." He shared his underlying work: gel electrophoresis, Sanger sequencing runs with quality scores, and multiple-sequence alignment (via the MAFFT server) matching his samples against SV40 and BNT162b2 reference sequences at high identity across several distinct plasmid features — the SV40 promoter, enhancer, origin, and poly(A) signal, plus the AmpR promoter and f1 origin common to the vector backbone. He also cross-checked the same signature against unrelated public cancer-genomics datasets as a specificity control. This is unpublished, preliminary work, shared publicly by its author — not yet a peer-reviewed paper — but it is primary laboratory data, independent of both Chakraborty's and McKernan's reanalyses, and it lands on the same vial/sample-level finding: the vector sequence is there. Louwen was explicit about the boundary of his own claim: "alleen contaminatie vraagstuk in sequencing data blijft bestaan" — the sequencing-data contamination question remains open. He is not claiming to have resolved it, only to have independently detected what the others detected, by a different method.
What hasn't changed, again: three separate people, three different methods (public-data reanalysis, public-data reanalysis, and now primary wet-lab sequencing), converge on the same detection-level finding — vector DNA sequence is present, more consistently than a pure sequencing artifact would predict. None of the three has shown it integrates into a genome. That line hasn't moved.
UPDATE 3 (13 August 2026) — Sentinel Protocol Expanded, One Bioavailability Claim Corrected
The cGAS-STING step of the Genomic Sentinel Protocol has been expanded with two additions, and one candidate compound was checked and left out.
Baicalin/baicalein (Scutellaria baicalensis, Chinese Skullcap) was added alongside resveratrol. Published research shows baicalein directly blocks phosphorylation of both STING and TBK1, and drives a structural liquid-to-solid phase transition in cGAS that shuts the sensor down rather than just dampening its output downstream — the most specifically-targeted mechanism of any compound in the stack.
EGCG (green tea extract) was added alongside curcumin and berberine. Its mechanism sits further upstream than either: it blocks G3BP1, the protein cGAS needs to bind cytoplasmic DNA in the first place, rather than inhibiting NF-κB after the alarm has already fired. It carries a real safety ceiling — EFSA's 2018 opinion links intake above 800mg/day to liver injury risk in concentrated supplement form, now flagged directly on the protocol page.
Rosmarinic acid (rosemary, lemon balm, oregano) was proposed as a third addition on the strength of a bioavailability claim — that it absorbs efficiently without needing piperine or fat, unlike curcumin. That claim didn't hold up: independent pharmacokinetic studies in humans and animal models put oral absorption at up to 1.69%, in the same poor range as unenhanced curcumin, not meaningfully better. Its NF-κB mechanism is also redundant with curcumin and berberine already in the stack. It was left out for lacking a distinct mechanism or a real absorption advantage — not for having a wrong mechanism.
What's Actually Solid, Independent of Any of This
Strip away the contested DNA-integration question and one thing survives on its own evidence, already documented in multiple peer-reviewed papers unrelated to this letter: vaccine mRNA itself circulates in blood for days to weeks, not hours, and not only at the injection site. Spike protein was detected in plasma in 3 of 13 participants an average of 15 days after the first dose2. A separate study found vaccine mRNA sequences circulating up to 28 days post-vaccination3. As late as September 2022, the American Red Cross was publicly stating that COVID-19 vaccines "do not enter the bloodstream" — a claim these papers, published before that statement, already contradicted.
That's the part of this story that doesn't need a contested reanalysis to stand up. It changes the practical picture regardless of what turns out to be true about the DNA-vector question: injected material reaching systemic circulation, for weeks, is the more basic and better-supported claim underneath the more dramatic one.
Where This Leaves the Terrain
You cannot build a botanical protocol around reversing a DNA integration event that hasn't been demonstrated to occur. No herb undoes a genomic sequence. What you can build a protocol around is what's true regardless of whether integration ever happened: the immune system has a dedicated sensor for foreign DNA in the wrong place — the cGAS-STING pathway — and it fires an inflammatory cascade whether the DNA is a manufacturing contaminant, a sequencing artifact caught in someone's data, or something more. Separately, the cell's ordinary genomic surveillance and DNA-repair machinery is worth strengthening on general principle, for anyone concerned about cumulative oncogenic risk from any source, not only this one.
That's the actual, honest scope of what's actionable here — laid out with its own methodological caveats in the Genomic Sentinel Protocol.
- McKernan K, Helbert Y, Kane LT, McLaughlin S (2023). Sequencing of bivalent Moderna and Pfizer mRNA vaccines reveals nanogram to microgram quantities of expression vector dsDNA per dose. OSF Preprints, April 10.
- Ogata AF, Cheng CA, Desjardins M, et al. (2022). Circulating SARS-CoV-2 vaccine antigen detected in the plasma of mRNA-1273 vaccine recipients. Clinical Infectious Diseases 74: 715–718.
- Castruita JAS, Schneider UV, Mollerup S, et al. (2023). SARS-CoV-2 spike mRNA vaccine sequences circulate in blood up to 28 days after COVID-19 vaccination. APMIS.
- Chakraborty S. The bloodstream of mRNA vaccinated individuals (both Pfizer and Moderna) shows DNA expression vector contamination, including SV40 and kanamycin-resistant gene sequences. Unpublished letter, no journal, no DOI.
- Vermeer J, Louwen R (2026). What We Found in the Earliest COVID-19 Patient Data. Submitted to Nature, July 2026. See the full investigation.
- Speicher DJ, Rose J, McKernan K (2025). Quantification of residual plasmid DNA and SV40 promoter-enhancer sequences in Pfizer/BioNTech and Moderna modRNA COVID-19 vaccines from Ontario, Canada. Autoimmunity 58(1):2551517. DOI: 10.1080/08916934.2025.2551517.