This article synthesises two evidentiary tracks that are both public and both still moving — a bioinformatic manuscript with open data, and an accountability process working through the US courts. It will be updated as verified material warrants, every addition dated.
Beyond Reasonable Doubt
For four years the origin of SARS-CoV-2 was presented to the public as an open question: two hypotheses, roughly balanced, awaiting data. That framing has not survived the documents. On 18 June 2024 Richard Ebright — a molecular biologist who has advised the US government on biosafety and select-agent policy for two decades — told a Senate committee the question was, for practical purposes, settled. After early drafts of the 2018 DEFUSE proposal were released under freedom-of-information law, his assessment narrowed to one sentence: there is no remaining room for reasonable doubt. This article is not another attempt to weigh two hypotheses against each other. It lays out the two separate bodies of evidence that closed the question — one recovered directly from the earliest patient samples, one assembled from six years of subpoenaed emails, FOIA releases and sworn testimony — and states the conclusion they support. It does not argue the release was deliberate, and it does not argue it was an accident. Where the virus came from and why it left the building are different questions. Only the first is answered here.
1. What Is in the Data
The first body of evidence does not depend on anyone's testimony. It is in the sequencing files themselves — the raw reads deposited by the Wuhan Institute of Virology behind Zhou et al. (2020), the paper that first described the outbreak, and in the published genome that every laboratory on earth has had since January 2020.
Seven Anomalies in the Earliest Patient Data
1. RBD-Fc clonality — the receptor-binding domain of the earliest patient sequences is 100% nucleotide-identical across 110 of 112 assemblies, against an expected ~0.17 mutations per copy under normal viral mutation rates over the claimed months of circulation, and matches the architecture of a patent (CN111333704B) filed 62 days after the sample date.
2. Undisclosed CRISPR/SpCas9 experiment — guide-RNA scaffold reads in one library (WIV05), with spacers matching five mouse genes (Cd68, Sat1, Mecp2, Procr, and a chrX region) at 100% identity, absent from every MiSeq control run.
3. GPS metadata anomaly — 101 sample records, submitted as one batch in January 2021, carry coordinates for Walter Reed National Military Medical Center, Bethesda, while the text field for each still reads "China: Wuhan."
4. Star phylogeny — 83 of 115 internal branches across 111 sequenced assemblies have lengths statistically indistinguishable from zero; calibrated against the nearest wild bat coronaviruses (ZC45/ZXC21), the branch lengths are 100 to 1,000 times shorter than genuine evolutionary distance in this region.
5. Restriction-enzyme signatures — the two spike insertions unique to SARS-CoV-2 (Insert 1, and Insert 4, which contains the furin cleavage site) carry flanking NcoI and dual NdeI sites consistent with directional molecular cloning; the two insertions SARS-CoV-2 shares with its nearest bat relative, RaTG13, carry no such signature.
6. Undisclosed co-processing of high-consequence pathogens — avian influenza H7N9 at 1,472× mean depth and 100% breadth in one library (WIV07-2), and Nipah virus carrying infectious-clone markers (an HDV ribozyme and a T7 terminator), present in two libraries and entirely absent — zero reads across 134 million — from a third library on the identical platform, which rules out simple background contamination.
7. Platform mislabeling — four of five deep-sequenced libraries are recorded in NCBI's metadata as "Illumina HiSeq," while file naming and depositor documentation confirm they were run on a different instrument (MGISEQ-2000RS).
None of these seven appears in the methods or the metadata of the paper that deposited the data. Each is independently reproducible from public files. Read together, they describe undisclosed synthetic-construct-grade laboratory material moving through the same pipeline that produced the first officially recognised COVID-19 patient samples.
Other researchers, the published genome, the same signature
The forensic layer is not one group's finding. Three independent lines of work — different people, different datasets, different years — arrive at the same place.
The Endonuclease Fingerprint
Working only from the published SARS-CoV-2 genome — no access to raw reads — this group mapped the positions of BsaI and BsmBI recognition sites, the two type IIS enzymes used for the standard "golden gate" method of assembling a coronavirus genome from segments. They reported that the sites are spaced to divide the genome into fragments of a length ideal for that assembly method, and that the longest fragment sits just under the practical size ceiling — a pattern they argued is vanishingly rare in natural sarbecoviruses and is exactly what an in-vitro assembly design produces. The paper was disputed on statistical grounds by other groups; the dispute is genuine and unresolved. What matters for this article is that an entirely separate method, applied to the public genome alone, points at the same category of origin as Finding 5.
Nipah Vector Sequences — Found Independently, Years Earlier
The Nipah half of Finding 6 was reported five years before the Vermeer & Louwen manuscript, by a different team screening the same WIV-deposited COVID-19 patient data. They found Nipah virus sequences carrying vector-cloning markers and flagged them as evidence of undisclosed infectious-clone work in the same libraries as the patient samples. Two independent screenings, half a decade apart, recovered the same reads.
The Same Institution, a Different Dataset, the Same Kind of Contaminant
A separate WIV-associated dataset — public agricultural rice sequencing data, nothing to do with the outbreak samples — was found by another team to contain a Merbecovirus infectious clone built on a pBAC-CMV backbone traced to Zhengli Shi's own laboratory. Different project, different year, different researchers: the same pattern of undisclosed synthetic-construct work moving through WIV sequencing pipelines without appearing in any published methods section.
The twelve nucleotides at the centre of it
SARS-CoV-2 carries a twelve-nucleotide insertion at the exact S1/S2 boundary of the spike gene that creates a furin cleavage site — the single feature most responsible for how efficiently the virus infects human cells, and the feature no other sarbecovirus has. The insertion is out of frame relative to the surrounding sequence. It is encoded by a codon pair (CGG-CGG) that appears in essentially no other coronavirus of this group and is rare across the genus. It sits precisely where a designed insertion would sit, and it alters the local restriction-site pattern in the way Finding 5 describes. Six years of examination by every research group with an interest, on all sides of the question, has not produced a natural sarbecovirus with this site or a documented evolutionary path to it.
2. The Timeline
The connections Ebright draws are not a web of names. They are a sequence, and the sequence is the argument: a written plan, the funding to execute it, the execution, and then a virus with the plan's exact features appearing where the work would have been done. Laid out in order — hard-dated facts in black, contested or attributed points marked.
3. What Is in the Paper Trail
The dated points above rest on a documentary record — not on re-analysed data. It came out of courts, congressional subpoenas and freedom-of-information litigation, mostly between 2023 and 2026, produced by people with no connection to the bioinformatic work and no knowledge of it. The four documents that carry the most weight:
DEFUSE: A Written Plan for a Virus With These Features
In 2018 a consortium led by EcoHealth Alliance — with the Wuhan Institute of Virology, Ralph Baric's laboratory at UNC, and Rocky Mountain Laboratories' Vincent Munster as a named partner — submitted DEFUSE, a roughly $14 million proposal to DARPA. The formal proposal was rejected for gain-of-function and dual-use risk. The early drafts and planning notes, obtained under FOIA by U.S. Right to Know and released in January 2024, are more specific than the submitted version. They describe: inserting furin cleavage sites at the S1/S2 junction of bat coronavirus spikes; assembling synthetic genomes from six DNA segments cut and rejoined with type IIS restriction enzymes; working with backbones up to 25% divergent from known SARS viruses; and selecting for receptor-binding domains adapted to human receptors. The drafts contain a line-item price quote for the BsmBI enzyme. They also show the WIV would carry out much of the work — a fact the drafters removed from the submission, in their own words, to make DARPA "comfortable."
Every specific design choice in that list corresponds to a feature of SARS-CoV-2 or to a finding in Section 1. Ebright, on the enzyme quote: "an order line for BsmBI in a draft of EcoHealth's 2018 DARPA proposal is the equivalent of a smoking gun."
The honest limit of this document: as microbiologist Alina Chan noted, the drafts are from early 2018. They prove intent, capability and a detailed blueprint within this exact network. They are not, by themselves, a record of the assembled virus or of a release. They are the plan; Section 1 is the residue.
"But the Proposal Was Rejected"
The most common reply to the DEFUSE evidence is that DARPA turned the proposal down, so the work in it never happened. Ebright's answer has three parts, and they are sound.
A rejected proposal is not a cancelled experiment. Submitting the same planned work to several funders at once is ordinary practice in the molecular life sciences — you write NIH, DARPA, a foundation, and proceed with whichever one comes through. A rejection tells you which cheque was not written. It does not tell you the research was abandoned.
The other channel was open. DEFUSE was the DARPA proposal. The NIH grant to EcoHealth covering the same category of work — gain-of-function on SARS-related coronaviruses — was not rejected. It was renewed in 2019, without the P3CO review the framework required, and it carried more than enough funding to perform the steps DEFUSE described. One door closed; the one next to it stayed open.
The result is itself evidence the work was done. Within roughly a year of that renewal, a virus carrying the specific combination of features set out in the 2018 proposals appeared next to the institute that would have carried them out. You do not need the lab notebook when the blueprint and the finished object match point for point.
"Would Leave No Signatures of Purposeful Human Manipulation"
A federal grant proposal co-authored by Baric's collaborator Fang Li described using "infectious clone technology" to test which coronaviruses could spill into humans. An FBI memorandum written weeks into the pandemic, later released under FOIA, quoted the proposal and drew the implication directly: the strategy, "if performed using commercial or in-house gene synthesis to prepare the infectious clones, would leave no signatures of purposeful human manipulation." The specific technique capable of producing a subtle, statistically-detectable-but-not-obvious signature — the kind documented in Finding 5 and by Bruttel et al. — was a funded, described method in this same network before the outbreak.
The Consensus Paper, in Its Authors' Own Words
"The Proximal Origin of SARS-CoV-2" (Nature Medicine, 17 March 2020) is the paper cited worldwide as evidence against a laboratory origin. The record now shows: Fauci personally proposed it to lead author Kristian Andersen on two days in early 2020, with same-day email evidence; Nature rejected an earlier draft for not arguing hard enough against a lab origin, and the paper's anchoring sentence — "we do not believe that any type of laboratory-based scenario is plausible" — was added afterward to satisfy a reviewer, confirmed under oath by Andersen and Garry; and in a Slack exchange months after publication Andersen told his co-authors "we really don't have any hard evidence one way or the other… we can't rule out that somebody actually put it in there." Two Erasmus MC gain-of-function researchers, Ron Fouchier and Marion Koopmans, contributed to the paper and declined to be named because they "opposed scientific articles considering the lab leak theory at all." Koopmans then joined the WHO origins team that called a lab origin "extremely unlikely." The full account is in The Pattern Behind the Pattern and The Rotterdam School.
The Pardon That Names a Date
The pre-emptive pardon issued to Anthony Fauci on the last day of the Biden administration covers conduct back to 1 January 2014. Ebright's observation: that is not an arbitrary boundary. It is the start date of the NIH grant to EcoHealth Alliance that funded bat-coronavirus research at the Wuhan Institute of Virology. A pardon defines the period it is written to protect.
The Accountability Process Is Now Moving
January 2025: HHS formally debarred EcoHealth Alliance and Peter Daszak for five years, citing failure to report gain-of-function experiments at the WIV. June 2026: the ODNI declassified intelligence-community records showing Lawrence Livermore National Laboratory had assessed a lab origin as equally likely as natural spillover in May 2020 — an assessment buried for six years — and that analysts who documented lab-origin scenarios faced career threats. 29 July 2026: Fauci invoked the Fifth Amendment rather than answer a Senate committee's questions about pre-pandemic gain-of-function funding. August 2026: David Morens — Fauci's senior adviser at NIAID for sixteen years — pleaded guilty to one count of conspiracy, having routed COVID-origins and terminated-grant emails through a personal account to defeat FOIA; sentencing is set for 12 November 2026. A DOJ criminal grand jury on COVID origins has been empanelled. Ebright's reading of the Morens plea: it is "possible, even likely," that Morens has agreed to testify against others.
4. What the Two Layers Together Establish
The data layer and the document layer were built by different people, using different methods, at different times, with no coordination between them. A bioinformatician screening public FASTQ files in 2026 was not talking to the FOIA litigators who pried loose the DEFUSE drafts in 2024, or to the congressional staff who took Andersen's deposition in 2023. And yet the two describe the same object from opposite ends.
Ebright calls it a one-for-one match: a blueprint for a crime, then the crime. Every design element written into the 2018 proposals corresponds to a feature the virus turned out to have.
| Written into the 2018 proposals (NIH + DARPA) | Found in SARS-CoV-2 |
|---|---|
| Spike gene up to ~25% divergent from SARS-CoV-1 | Spike ~24% divergent from SARS-CoV-1 |
| Selected for high binding affinity to the human receptor | Exceptionally high human-ACE2 binding affinity |
| A furin cleavage site added at the S1/S2 junction | A furin cleavage site at the S1/S2 junction — carried by no other sarbecovirus among 800+ sequenced |
| Genome assembled from six fragments joined with BsmBI | Restriction-site spacing consistent with six-fragment assembly (Finding 5; Bruttel et al.) |
| Work to be performed at the Wuhan Institute of Virology | Outbreak began in Wuhan, >1,000 km from the nearest wild reservoir, at that institute's doorstep |
Each row on its own could be argued as coincidence. The furin cleavage site alone: none of the 800-plus sequenced SARS-related coronaviruses carries one — so the base rate of finding it in a member of this group by chance is under one in eight hundred, before accounting for the fact that it sits at the exact junction a proposal named a year earlier, in the exact city that proposal named. Stack that against the divergence range, the receptor affinity, the six-fragment signature and the location, and coincidence stops being an available explanation.
The proposal describes assembling a SARS-like virus from six restriction-enzyme fragments and adding a furin cleavage site at S1/S2. The published virus has a furin cleavage site at S1/S2 that no relative shares, and a restriction-site pattern spaced for six-fragment assembly. The earliest patient samples carry that virus alongside undisclosed synthetic-construct material, a patent-matched receptor-binding domain, CRISPR reagents and infectious-clone markers for other pathogens. The paper that declared the origin natural was solicited by the official who funded the network, rewritten to strengthen that conclusion at a reviewer's request, and privately doubted by its own lead author. The competing hypothesis — emergence at the Huanan market — has had six years and more than eighty thousand animal samples and has not produced a single infected intermediate host, and the earliest known cases are not the market cases.
On the standard of evidence used in every other empirical field — multiple independent lines converging, no alternative that accounts for all of them — this is not a balanced question. It has an answer.
The conclusion, stated plainly
SARS-CoV-2 originated in laboratory work. The forensic data recovered from the earliest patient samples, the independent analyses of the published genome, and the documentary record released since 2023 are three separate bodies of evidence, and they converge. There is no natural-origin account that explains the furin cleavage site, the restriction-site architecture, the clonal receptor-binding domain, the patent match, and the undisclosed co-processing of engineered pathogens together. There is a laboratory account that explains all of them, and it was written down, in this network, in 2018.
What this does not say: it does not say the release was deliberate, and it does not say it was an accident. The evidence for where the virus came from is not evidence of intent. That is a separate question, it is not answered by anything above, and this article takes no position on it. Conflating the two is how the origin question was kept open for six years — treat them separately.
Sources
- Vermeer, J. & Louwen, R. "Seven Anomalies in the Earliest SARS-CoV-2 Sequencing Data: An Independent Bioinformatic Analysis of PRJNA605983." Zenodo, DOI 10.5281/zenodo.21410252, 2026; submitted to Nature as a formal Correction/Retraction request. Data and pipelines: github.com/jasper-cmyk/PRJNA605983-analysis.
- Bruttel, V., Washburne, A. & VanDongen, J. "Endonuclease fingerprint indicates a synthetic origin of SARS-CoV-2." bioRxiv 2022.10.18.512756, 20 October 2022. See also the published rebuttals (e.g. PMC9937728) — the statistical dispute is unresolved.
- Quay, S., Zhang, D., Jones, A. & Deigin, Y. "Nipah virus vector sequences in COVID-19 patient samples sequenced by the Wuhan Institute of Virology." arXiv:2109.09112, 2021.
- Jones, A., Zhang, D., Massey, S. & Nemzer, L. "Discovery of a novel merbecovirus DNA clone contaminating agricultural rice sequencing datasets from Wuhan, China." Journal of Bioinformatics and Systems Biology, January 2024.
- DEFUSE proposal (DARPA HR001118S0017, 2018) and early drafts released under FOIA by U.S. Right to Know, January 2024. Reporting: "New Documents Bolster Lab-Leak Hypothesis," City Journal, January 2024; usrtk.org.
- Written Testimony of Richard H. Ebright, US Senate Homeland Security and Governmental Affairs Committee, 18 June 2024, hsgac.senate.gov.
- FBI internal memorandum, "[Suspect] Re: Follow up call," 23 April 2020, released via FOIA.
- Select Subcommittee on the Coronavirus Pandemic, "The Proximal Origin of a Cover-Up: Did the 'Bethesda Boys' Downplay a Lab Leak?", Majority Staff Report, 11 July 2023, oversight.house.gov.
- Slack export "slack-drop-pm.pdf," released by the Senate Homeland Security and Governmental Affairs Committee, paul.senate.gov/readingroom, 21 July 2026.
- Andersen, K., Rambaut, A., Lipkin, W.I., Holmes, E. & Garry, R. "The Proximal Origin of SARS-CoV-2." Nature Medicine, 17 March 2020.
- Presidential pre-emptive pardon of Anthony S. Fauci, 20 January 2025 (covering conduct from 1 January 2014).
- HHS debarment of EcoHealth Alliance and Peter Daszak, January 2025.
- ODNI Press Release PR-11-26, "Director Gabbard Releases Declassified COVID-19 Origins Materials," 18 June 2026, odni.gov.
- US Senate hearing on COVID-19 origins, 29 July 2026 (Fauci invocation of the Fifth Amendment).
- United States v. David Morens, District of Columbia — indictment April 2026; plea agreement signed 20 July 2026; guilty plea to one count of conspiracy, August 2026; sentencing scheduled 12 November 2026.
- Richard Ebright, remarks to Blaze News on the Morens plea, August 2026; and BlazeTV "Rufo & Lomez," 2026 (pardon start-date analysis).
- SureChEMBL patent record, CN111333704B / application CN-202010112679-A (Zhou Yusen et al., filed 24 February 2020).
The Reference
500 botanicals — each with documented mechanisms, dosages, and synergy protocols. The complete reference for sovereign health practice.
The 500 Most Powerful Botanicals →