Searching for RBD-Fc Junction Reads in the Earliest Wuhan Patient Samples

1. Introduction

1.1 Background

On 24 February 2020, PLA military researcher Brigadier-General Yusen Zhou filed patent CN111333704 describing an RBD-Fc fusion protein vaccine for SARS-CoV-2 — within weeks of the outbreak declaration. Given that Zhou's identical MERS-CoV Fc-fusion approach (2013) took a minimum of four months from conception to filing, a December 2024 report by Dr. Robert Kadlec (Texas A&M / Bush School) calculated that work on this construct began no later than November 2019, and possibly as early as summer 2019.

In June 2026, molecular biologist Dr. Rogier Louwen (@dr_louwen) reported finding reads matching this RBD-Fc construct in the public BALF sequencing data from PRJNA605983 — the earliest Wuhan patient samples. This followed independent work by Quay et al. (arXiv 2109.09112, 2021), which had already identified Nipah virus infectious clone sequences, pcDNA3.1+, and other laboratory constructs in the same dataset — a finding that has never been retracted or formally addressed.

1.2 The Critical Question

The key evidentiary threshold was articulated by @BillyBostickson on 18 June 2026: do any reads in this dataset span the RBD→Fc junction in a single continuous alignment? A junction-spanning read would place viral and synthetic sequences on the same RNA molecule — making library contamination an insufficient explanation. This report is a direct attempt to answer that question.

2. Methods

2.1 Data Acquisition

Data integrity confirmed: SRA version=1, no last_update field. MD5 checksums unchanged since original upload. Platform is listed as "Illumina HiSeq 3000" in SRA metadata, but original filenames (v300043428_L04_123_1.fq.gz) identify the instrument as MGISEQ-2000RS — a metadata discrepancy that does not affect alignment (identical Phred+33 encoding).

2.2 Reference Construction

The RBD-Fc junction reference was derived from patent CN111333704 SEQ ID NO.5 (1,173 bp after trimming). The junction between the SARS-CoV-2 RBD and the human IgG1 Fc domain was located bioinformatically at position 477 using the IgG1 hinge start motif GAACCTAAATCTTGTGAC (= EPKSCDKT).

Amino acid sequence at the boundary: ...Val-Asn-Gln-Ile-Asn-Ala-Asn | Ala-Ala-Ala-Glu-Pro-Lys-Ser-Cys-Asp...

This direct RBD→Fc fusion does not occur in nature. A 150 bp read must span positions 402–552 (≥75 bp each side) to qualify as junction-spanning.

Multi-reference FASTA for simultaneous Scenario A + B screening:

The H7 HA alignment was run as a separate pass against WIV07-2 following a query by Dr. Monali Rahalkar (@MonaRahalkar) about hemagglutinin reads in this sample, consistent with the pVAX1-HA finding in Quay et al. (2021).

2.3 Alignment

BWA-MEM naturally routes junction-spanning reads to the RBD-Fc reference: reads containing only RBD sequence prefer NC_045512.2 (longer, higher BWA score), while reads containing Fc sequence find no match on NC_045512.2 and align to the RBD-Fc reference. This provides an intrinsic filter without manual read separation.

2.4 Junction Detection — CIGAR-Correct Calculation

This correction collapsed 8 apparent junction reads in WIV05 to 0 after proper CIGAR parsing — the 8 reads each had CIGAR patterns like 103S20M27S, mapping only 20 bp near position 384, with the rest soft-clipped.

3. Results

3.1 Overall Alignment

Low mapping rate is expected: BALF RNA-seq is dominated by human transcriptome not present in our reference.

3.2 Scenario A — RBD-Fc Junction

Coverage profile (WIV05):

The step-change at position 477 is diagnostically unambiguous. Coverage left of the junction is background noise. Coverage right of the junction is the patient's own IgG1 antibody mRNA. The 2 reads nominally at position 477 in WIV05 have CIGAR 66S19M65S and 22S19M109S — 19 matched bases at the hinge entry motif, massive soft-clips. These are B-cell reads, not construct reads.

3.3 Scenario B — Synthetic Construct Signals

3.4 H7N9 Hemagglutinin Signal (WIV07-2)

A separate alignment of WIV07-2 reads against H7N9 HA (KC853766.1, A/Hangzhou/1/2013) and H7N2 HA (AF072399.3) was performed in parallel with the multi-reference run.

Per-base coverage ratio H7N9 HA / SARS-CoV-2: 16.8×. H7N9 HA has nearly 17 times deeper coverage per base than the virus this patient was hospitalized with.

Coverage is uniformly distributed across all 1,683 bp of the HA gene (918×–2,292× in 10 evenly spaced windows, no zero-coverage positions). This profile is inconsistent with cross-mapping artifact — which produces coverage hotspots, not a flat plateau — and inconsistent with active H7N9 patient infection, which would produce reads from all eight influenza genome segments, not only segment 4. The single-segment, high-coverage, uniform profile is the expected signature of a cloned HA insert in a high-copy expression vector such as pVAX1.

This independently replicates Quay, Rahalkar, Jones & Bahulikar (2021, Zenodo:5067706), who assembled a complete 4,765 bp pVAX1-H7N9-HA plasmid de novo from WIV07-2 reads using SPAdes, with ORF2 (274–1956 bp) showing 100% nucleotide identity to H7N9 HA (KF021597.1). Their K-mer analysis (fastv) yielded an HA/SARS-CoV-2 ratio of 600% in WIV07-2 — our BWA-MEM alignment yields a per-base coverage ratio of 1,472×/87.8× = 16.8×, consistent in direction and order of magnitude.

The H7N9 HA signal was first detected by Abouelkhair (2020, PeerJ), who analysed the same dataset using K-mer analysis and reported Influenza A (H7N9) among multiple non-SARS-CoV-2 sequences. Abouelkhair interpreted these findings as patient co-infections. Rahalkar, building on Abouelkhair's detection, reanalysed the data and identified the H7N9 HA as vectorized — cloned into pVAX1 — reaching a fundamentally different conclusion: laboratory contamination rather than clinical co-infection. The distinction is now resolved with our own MiSeq control data (Section 3.5): alignment of MiSeq WIV06 (SRR11092056) against KC853766.1 yields only 8 reads (0.76 RPM, 4.5% breadth, 0.09× depth) — LAHH background noise indistinguishable from zero. MGISEQ WIV07-2 yields 18,046 reads (234.6 RPM, 100% breadth, 1,472×). The 309× per-million read differential is incompatible with patient H7N9 co-infection, which would produce signal on both sequencing platforms.

A WIV publication documented that DNA vaccines containing H7N9 HA genes were being developed and tested in mice at WIV during 2019–2020. The co-occurrence of a pVAX1-H7N9-HA construct with early COVID-19 patient BALF samples processed at the same facility is the central forensic question identified by Quay, Rahalkar, Jones & Bahulikar (2021) and now independently confirmed here.

Nipah Bangladesh (AY988601): Most consistent cross-sample signal. 634–683 reads in both samples. Independently replicates Quay et al. (2021). Has never been addressed by the submitting institution.

SpyCas9 (KM099231): High read counts but concentrated in two hotspots (pos 1,500–2,000: up to 2,565×; pos 6,500–7,000: up to 6,635×). Non-uniform profile is inconsistent with intact plasmid presence. Follow-up solo alignment (SpyCas9-only reference, no competing references) confirmed cross-mapping origin: ColE1 ori region (pos 195–783): 100% breadth / 30.9× depth; AmpR CDS (pos 8911–9720): 36.9% breadth / 1.4× depth; SpyCas9 CDS (pos 2122–6396, Cas9-specific): 8.1% breadth / 0.3× depth — LAHH noise indistinguishable from zero. The apparent signal originates entirely from ColE1 ori and AmpR backbone elements shared with pcDNA3.1 (PZ020853.1). This reference is removed from the confirmed findings.

pcDNA3.1 / HPV-16 E6/E7 (PZ020853.1): The reference sequence is pcDNA3.1:(noATG)AU1-16E6starI_5'E7-GFP (6,463 bp) — a mammalian expression vector carrying HPV-16 E6 and E7 oncogene sequences with a GFP reporter. WIV07-2 MGISEQ shows 44,291 reads, 91.5% coverage breadth, and 877.9× average depth — the strongest absolute signal of any non-SARS-CoV-2 reference in this study, approaching the coverage breadth of SARS-CoV-2 itself (99.9%). Coverage peaks at positions 3,300–3,900 (up to 4,646×) correspond to HPV-16 E6/E7 sequences. The MiSeq platform control (Section 3.5) definitively resolves the co-infection interpretation: MiSeq WIV06 shows only 92 reads, 10.5% coverage breadth, and 0.3× average depth — LAHH background noise indistinguishable from zero. Patient HPV co-infection would produce signal on both sequencing platforms. It does not. The 8.7× MiSeq/MGISEQ breadth ratio for this vector is the largest platform differential measured in this study.

3.5 MiSeq Platform Control (SRR11092056, WIV06)

To directly test the machine-specific clustering claim (Bostickson & Ghannam, 2021, Section 13), we aligned the MiSeq run WIV06 (SRR11092056; 10,479,446 read-pairs, 150 bp) against the same multi-reference FASTA used for the MGISEQ samples. Coverage breadth — the fraction of a genome covered by at least one read — is the critical discriminating metric. Low Abundance High Homology (LAHH) cross-mapping noise produces scattered coverage in ≤20% of a genome at sub-1× average depth. Genuine biological signal produces >30% breadth with consistent multi-read coverage.

The SARS-CoV-2 breadth difference (45.2% MiSeq vs. 97.9–99.9% MGISEQ) reflects different patients and different viral loads between runs — not a platform artefact. Within each sample, non-SARS-CoV-2 signals should be evaluated relative to the SARS-CoV-2 baseline of that same run.

In MiSeq WIV06, the Nipah breadth (8.3%) is 18% of the SARS-CoV-2 baseline (45.2%) — consistent with LAHH scatter. In MGISEQ WIV05, the same ratio is 33% (32.0% / 97.9%). In MGISEQ WIV07-2, the HPV E6/E7 breadth (91.5%) reaches 92% of the SARS-CoV-2 baseline (99.9%) — near-equivalent coverage of a lab construct and the patient's primary pathogen.

The pCas9 vector (KM099231) was investigated via solo alignment (SpyCas9-only reference, removing competition from pcDNA3.1). Per-region breakdown: ColE1 ori (pos 195–783) 100% breadth / 30.9×; AmpR CDS (pos 8911–9720) 36.9% / 1.4×; SpyCas9 CDS (pos 2122–6396) 8.1% / 0.3×. The Cas9-specific coding region shows LAHH-level noise. The signal in multi-reference alignment originates entirely from backbone elements (ColE1 ori, AmpR) shared between pCas9 and pcDNA3.1. pCas9 is excluded from confirmed findings.

Library depth normalisation (reads per million)

Because MGISEQ samples contain 6–7× more total reads than MiSeq WIV06 (68–77M vs. 10.5M), a direct read-count comparison could be confounded by library depth. Reads per million (RPM) normalises for this:

H7N9 HA: MiSeq WIV06 shows 8 reads (0.76 RPM, 4.5% breadth, 0.09× depth) — absolute LAHH noise over a 1,683 bp gene. MGISEQ WIV07-2 shows 18,046 reads (234.6 RPM, 100% breadth, 1,472× depth). The platform differential is 309× per million reads, far exceeding what a 7.3× library size difference could produce. This is the largest RPM differential of the three confirmed signals.

Nipah: MiSeq WIV06 shows 24.0 RPM — 2.6× more per-million reads than MGISEQ WIV05 (9.3 RPM). Despite proportionally more Nipah reads, MiSeq covers only 8.3% of the genome vs. 32–43% in MGISEQ. A depth effect predicts higher breadth at higher RPM. The inverse is observed.

HPV-16 E6/E7 (WIV05): MiSeq 8.8 RPM vs. MGISEQ WIV05 10.7 RPM — a 1.2× difference in read rate. Coverage breadth: 10.5% vs. 75% — a 7.1× difference. Nearly matched per-million read rates, radically different breadths. This isolates the platform effect independently of library size.

HPV-16 E6/E7 (WIV07-2): 575.8 RPM vs. 8.8 RPM — 65× more per-million reads in MGISEQ. This 65-fold excess cannot be accounted for by the 7.3× depth difference; the signal is genuinely over-represented per sequenced read on the MGISEQ platform.

3.6 SpyCas9 gRNA Scaffold — Direct Read Detection (Four Samples)

The SpyCas9 gRNA scaffold is a 76 bp sequence common to all SpyCas9-compatible guide RNA expression constructs. Because this element is shorter than a single 150 bp paired-end read, BWA-MEM alignment against a 76 bp reference is not applicable. Instead, we searched for an unambiguous 30 bp internal substring of the canonical scaffold (GTTTTAGAGCTAGAAATAGCAAGTTAAA, forward and reverse complement) by direct string matching across the raw FASTQ files of all four sequenced samples: WIV04 (SRR11092057, MiSeq), WIV05 (SRR11092061, MGISEQ), WIV06 (SRR11092056, MiSeq), and WIV07-2 (SRR11092059, MGISEQ).

Every positive read shows the same three-element architecture: a 3′ tail of the human U6 RNA polymerase III promoter (GTGGAAAGGACGAAACACCG), followed by a unique 20 bp protospacer, followed by the complete 76 bp SpyCas9 gRNA scaffold. Representative reads:

SRR11092057.181859:
  ...GTGGAAAGGACGAAACACCG[CCAGTAGATAAGCGACTGTC]GTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGG

SRR11092057.3534218:
  ...GTGGAAAGGACGAAACACCG[CCGCCAGCATGCGCTGCCCG]GTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGT

[brackets indicate unique 20 bp spacer]

This three-component structure — U6 promoter → variable spacer → canonical gRNA scaffold — is the defining architecture of a pooled sgRNA library, a tool used in genome-wide CRISPR screens where thousands of distinct guide RNAs are expressed from a shared promoter cassette. The 11 distinct spacer sequences found across the two positive samples are unique to each read. The spacers found in WIV04 (CCAGTAGATAAGCGACTGTC, CCGCCAGCATGCGCTGCCCG) do not overlap with any spacer found in WIV06 — expected behaviour when randomly sampling trace quantities from a large pooled library.

Platform directionality: This signal follows the inverse platform pattern compared to the Nipah, H7N9 HA, and HPV-16 E6/E7 findings. Those three signals cluster in MGISEQ samples (Group 2); the gRNA scaffold reads cluster in MiSeq samples (Group 1, WIV04 and WIV06). The two patterns are not in conflict — they represent two independent contamination axes from two different sequencing contexts, each platform-specific in its own direction.

4. Discussion

The RBD-Fc junction read — a single 150 bp read containing ≥75 bp of SARS-CoV-2 RBD sequence connected without interruption to ≥75 bp of synthetic IgG1 Fc sequence — was not present in either WIV05 or WIV07-2. This does not prove the construct was never present in these patients. It means the available public data does not demonstrate it.

The Nipah Bangladesh signal — first identified by Chakraborty (2020) and subsequently documented by Quay et al. (2021), Zhang (2021, Zenodo:4486195), and Bostickson & Ghannam (2021) — remains the most scientifically significant finding. Nipah (BSL-4, CFR ~70%) has no plausible biological reason to appear in COVID-19 BALF samples unless the processing laboratory was simultaneously conducting Nipah work at BSL-4 level, with cross-contamination into library preparation. Zhang (2021) provided the most granular structural evidence: the Nipah sequences carry a 3'-HDV ribozyme, T7 terminator, and tetracycline resistance marker — three independent markers of an assembled infectious clone construct, not natural viral RNA fragments. Bostickson & Ghannam (2021) further showed that the Nipah signal is machine-specific: it appears exclusively in the HiSeq 3000 runs (Group 2: WIV02-2, WIV04-2, WIV05, WIV06-2, WIV07-2), not in the MiSeq runs of the same patients — a pattern characteristic of sequencing centre cross-contamination rather than patient co-infection. This has not been addressed by the submitting institution in over five years.

The H7N9 HA signal in WIV07-2 adds a second independent line of evidence. The 16.8× per-base depth advantage over SARS-CoV-2, combined with 100% uniform coverage and absence of other influenza segments, is consistent with a pVAX1-HA expression construct — a conclusion reached independently by Quay, Rahalkar, Jones & Bahulikar (2021), who assembled the complete plasmid de novo, and by Zhang (2021, Zenodo:4486195), who identified the H7N9 HA under a CMV promoter with bgH poly(A) in SRR11092059–62 — the same sample subset, analysed independently with different methods, reaching the same vectorized-construct conclusion. A WIV publication confirmed active H7N9 HA vaccine development at the same institute during 2019–2020.

On the co-infection vs. contamination dispute: Abouelkhair (2020) published the first detection of these non-SARS-CoV-2 sequences in PeerJ (peer-reviewed) and interpreted them as evidence of patient co-infections. This interpretation was later challenged by Rahalkar et al. (2021), who found the H7N9 HA embedded in a pVAX1 expression vector — a configuration that cannot arise from a natural influenza infection. The machine-specific clustering (Group 2 MGISEQ only, not Group 1 MiSeq) was reported by Bostickson & Ghannam (2021, Section 13) as the decisive discriminating evidence. We have now reproduced this result directly (Section 3.5): MiSeq alignment of WIV06 (SRR11092056) against the same multi-reference FASTA yields only LAHH-level coverage for Nipah (8.3% breadth, 0.27× depth) and HPV-16 E6/E7 (10.5% breadth, 0.3× depth), while MGISEQ samples show 32–43% breadth for Nipah and 75–91.5% breadth / up to 877.9× depth for HPV E6/E7. This 4–8× platform differential for non-SARS-CoV-2 sequences, absent for SARS-CoV-2 itself, is incompatible with patient co-infection — which would produce signal regardless of sequencing platform — and is consistent with cross-contamination introduced at the sequencing facility during library preparation.

Quay, Rahalkar, Jones & Bahulikar (2021) also documented a third contamination layer: Spodoptera frugigiperda rhabdovirus (SfRV) at 13–83% of SARS-CoV-2 read levels in all five MGISEQ samples. SfRV is endemic to Sf9 insect cell lines, which are the standard host for baculovirus-based protein expression systems used in vaccine manufacturing. Its consistent presence across all five patients reinforces the picture of a sequencing facility simultaneously processing patient material and insect-cell-derived vaccine constructs.

Limitations

1. Reference accuracy: If the actual construct used different codon optimization or Fc variant than CN111333704 SEQ5, alignment sensitivity would be reduced.
2. pcDNA3.1 reference: PZ020853.1 contains HPV-16 E6/E7 inserts. Machine-specific clustering confirms laboratory contamination, but a clean pcDNA3.1 backbone reference would allow separating vector backbone signal from HPV-16 gene signal.
3. Sample scope: Primary analysis on WIV05 and WIV07-2; MiSeq control on WIV06 (SRR11092056). Three additional MGISEQ runs (SRR11092060, SRR11092062, SRR11092063) not examined.
4. RNA-seq scope: Plasmid DNA not actively transcribed would not be detected by this method, despite the CMV promoter in pcDNA3.1.

5. Conclusion

6. Data Availability & Reproducibility

All steps are reproducible using standard SRA toolkit and BWA on any Linux system with NCBI access. No proprietary software or institutional access required.

7. References

1. Zhou Y et al. (2020). Patent CN111333704 — Recombinant protein based on SARS-CoV-2 receptor binding domain. Filed 24 February 2020.
2. Quay SC^#, Rahalkar MC^#, Jones A^#, Bahulikar R. (2021). "Contamination or Vaccine Research? RNA Sequencing data of early COVID-19 patient samples show abnormal presence of vectorized H7N9 hemagglutinin segment." Zenodo. DOI:10.5281/zenodo.5067706. [Equal authorship: Quay, Rahalkar, Jones]
3. Wu F et al. (2020). "A new coronavirus associated with human respiratory disease in China." Nature 579:265–269. BioProject PRJNA605983.
4. Kadlec RP. (2024). "Hidden in Plain Sight — Muddy Waters Update: First Installment." Bush School, Texas A&M University.
5. Li H, Durbin R. (2009). "Fast and accurate short read alignment with Burrows-Wheeler Aligner." Bioinformatics 25(14):1754–1760.
6. Danecek P et al. (2021). "Twelve years of SAMtools and BCFtools." GigaScience 10(2):giab008.
7. Chakraborty S. (2020). First identification of Bangladeshi Nipah strain contamination in early Wuhan COVID-19 patient BALF samples (PRJNA605983). [Preprint / independent analysis, 2020].
8. Abouelkhair MA. (2020). "Non-SARS-CoV-2 genome sequences identified in clinical samples from COVID-19 infected patients: Evidence for co-infections." PeerJ 8:e10246. DOI:10.7717/peerj.10246. PMC7643552. [Interpreted findings as patient co-infections; subsequent machine-specific clustering analysis by Rahalkar et al. (2021) supports contamination over co-infection.]
9. Bostickson W, Ghannam Y. (2021). "Proposed Forensic Investigation of Wuhan Laboratories." Section 13: Illumina Machine Cross Contamination. ResearchGate. DOI:10.13140/RG.2.2.11804.26245.
10. Rahalkar MC. (2026). Personal communication via X/@MonaRahalkar, 23 June 2026: confirmed H7N9 HA discovery was initiated after Abouelkhair (2020), leading to collaboration with Quay et al. (Zenodo:5067706).
11. Zhang D. (2021). "Vector sequences in early WIV SRA sequencing data of SARS-CoV-2 inform on a potential large-scale security breach at the beginning of the COVID-19 pandemic." Zenodo. DOI:10.5281/zenodo.4486195. [Independent analysis of SRR11092059–62 (PRJNA605983); identified Nipah with 3'-HDV ribozyme + T7 terminator + tetracycline resistance (infectious clone markers) and H7N9 HA under CMV promoter + bgH poly(A) (= pVAX1 architecture). Prior independent work, completed before our analysis.]

This analysis was conducted independently and is not affiliated with any institution. No financial interest in any outcome of the COVID-19 origins debate.
Correspondence: jasper@sovereignhealthbotanicals.com