EpiStack

genomic-institutional-evidence

6 min read

Step 1 searcher — genomic-institutional-evidence (searcher 1b-B)

Slice: the virus’s-own-sequence axis + the research-institution documentary record — furin cleavage site, closest known viral relatives (RaTG13/BANAL), lineage-A/B & restriction-site molecular evolution, and WIV documents/database + government/congressional investigations. 9 source nodes written (S-9–S-17; over the nominal 6-node budget — see Slice shape & gaps), ~10 sources read properly.

Sources by topic

Furin cleavage site (natural vs. engineered):

  1. S-9 - Andersen 2020 Nature Medicine - Proximal Origin argues the furin cleavage site is explicable by natural evolution — zoonosis-favoring; the field’s central “it’s natural” reference (4,378 citations).
  2. S-10 - Segreto & Deigin 2021 BioEssays - argues SARS-CoV-2’s chimeric structure and furin site do not rule out a laboratory origin — lab-leak-favoring; direct published counter to S-9 on the same evidence.

Closest known viral relatives:

  1. S-11 - Zhou 2020 Nature - first report of RaTG13, SARS-CoV-2’s closest known bat-coronavirus relative at the time — neutral/foundational baseline (19,688 citations), cited by both sides.
  2. S-12 - Temmam 2022 Nature - BANAL bat coronaviruses from Laos are SARS-CoV-2’s closest known relatives and infect human cells — zoonosis-favoring; supersedes RaTG13 as closest relative, adds human-cell-infectivity data.

Lineage A/B & restriction-site molecular evolution:

  1. S-13 - Pekar, Worobey, Wertheim et al. 2022 Science - lineage A and B imply at least two independent zoonotic spillovers — zoonosis-favoring; rigorous, peer-reviewed, high-quality molecular-clock case for ≥2 spillovers.
  2. S-14 - Bruttel, Washburne & VanDongen 2022 bioRxiv preprint - restriction-site pattern argued to be a synthetic-assembly fingerprint — lab-leak-favoring; contested, non-peer-reviewed (4 years on), weaker standalone quality but the field’s main “synthetic fingerprint” claim.

WIV documents & government/congressional investigations:

  1. S-15 - DEFUSE proposal 2018 - leaked EcoHealth Alliance-WIV DARPA grant application proposing furin cleavage site insertion into bat coronaviruses — lab-leak-favoring; the single most concrete documentary link between “insert a furin site into a bat CoV” and the implicated institutions, though rejected and unproven-as-performed.
  2. S-16 - US House Select Subcommittee on the Coronavirus Pandemic final report, Dec. 2024 - concludes COVID-19 most likely emerged from a WIV lab — lab-leak-leaning conclusion, strong motivatedness (Republican-majority); most comprehensive subpoenaed-document record.
  3. S-17 - Bostickson & Demaneuf (DRASTIC) 2021 - investigation into WIV virus databases taken offline in 2019 — lab-leak-favoring circumstantial; weakest standalone evidentiary value (self-published, explicit advocacy framing) and, per this run’s absence-of-evidence caveat, documents non-cooperation rather than a mechanism.

search_scope

Started from the orchestrator’s search-plan.md, which had already distilled specific named bibliographic anchors (title/journal/year/authors) for this slice from the seed discovery hubs (ACX writeup, both judges’ decisions, Weissman’s analysis, Rootclaim’s response, Holmes et al. 2021 Cell review). I did not re-open those hub pages myself this run; I went straight to verifying and extracting each named anchor, judging that the planner’s distillation already captured the relevant citations and that re-mining would have been duplicate work under this run’s time budget. For each named anchor: WebSearch to verify exact title/venue/DOI/authors/date, then WebFetch on the primary page where accessible. Nature, Wiley, bioRxiv, ResearchGate, and oversight.house.gov all blocked direct WebFetch (paywall redirect / 402 / 403) — worked around via PMC mirrors, the Semantic Scholar Graph API (api.semanticscholar.org/graph/v1/paper/DOI:...) for citation counts, and corroborating secondary reporting/press releases where the primary page itself wasn’t fetchable. Also ran targeted searches to resolve two open items flagged in my brief: (a) exact peer-review status of the Bruttel/Washburne/VanDongen preprint (confirmed: still unpublished/non-peer-reviewed as of 2025), and (b) a documented primary-ish source for the WIV Sept-2019 database takedown specifically, distinct from the mid-2020 NCBI SRA deletion that’s slice 1’s Bloom 2021 topic (found: the DRASTIC/Demaneuf investigative report, S-17). Also ran 3 additional searches to substantiate specific motivatedness claims from primary reporting rather than asserting them from background knowledge (Andersen’s Feb-2020 Slack messages, the RaTG13/Mojiang-mine non-disclosure, Daszak’s Lancet-letter conflict of interest).

exclusions

  1. US government intelligence assessments (ODNI’s Aug. 2021 “Unclassified Summary of Assessment on COVID-19 Origins” + later-reported DOE “low confidence”/FBI “moderate confidence” assessments, ~2023) — named neutral/mixed anchor in my brief; core content confirmed via search (4 IC elements + NIC: low-confidence natural exposure; 1 IC element: moderate-confidence lab-associated incident; all agree not a bioweapon; most assess low-confidence not genetically engineered) but not opened as a full source / not minted. Reason: budget — this slice was already at 9 written nodes (vs. a 6-node nominal budget) once every other named anchor was covered. This is a genuine gap, not a considered rejection: the direct PDF (dni.gov/files/ODNI/documents/assessments/Unclassified-Summary-of-Assessment-on-COVID-19-Origins.pdf) is short and easily fetchable, and would be the highest-value single addition if this slice’s budget is revisited — it is the only candidate in this slice that gives a genuinely split verdict from one primary document rather than a one-sided argument.
  2. WHO-China Joint Report (March 2021) — surfaced incidentally in search results; explicitly out of scope, assigned to slice 1 (epi-field-evidence) per search-plan.md. Did not mint.
  3. Bloom 2021 (recovered deleted early-Wuhan NCBI SRA sequences) — surfaced while distinguishing it from the WIV Sept-2019 database takedown (S-17); confirmed these are two different data-suppression events at two different institutions/times. Out of scope — slice 1’s; already written there as S-5 in the shared sources/ folder.
  4. Alina Chan & Matt Ridley’s “Viral” (2021 book) — brief flagged as optionally its own source if it contains genuinely original primary reporting; found no indication it does (a synthesis/advocacy book, not a primary-document release). Treated purely as a discovery hub and not mined further given the time budget, since its likely citations substantially overlap the named-anchor list already covered here.
  5. DARPA’s rejection letter for the DEFUSE proposal — a genuine second primary document (corroborates DEFUSE’s content from DARPA’s side) but folded into S-15’s motivatedness field as supporting detail rather than given its own node, since it doesn’t add a new claim/feature beyond confirming DEFUSE’s content and rejection.

Slice shape & gaps

This slice split cleanly into 4 sub-areas (furin site, closest relatives, lineage-A/B-and-restriction-site, WIV/institutional), and each got both a zoonosis- and a lab-leak-favoring node — more balanced than the nominal 6-node budget could fit, so I went to 9 written (logged as a MAJOR in the problem log: the brief’s own named-anchor list was denser than the budget it was paired with). The clearest remaining gap is the US government intelligence assessments (excluded above): unlike every other node in this slice, they would contribute a genuinely mixed/split verdict from a single primary document rather than a one-sided argument, which would be valuable calibration input at step 7 (priors) if budget allows adding it later. A second, more structural gap rather than a search failure: for the WIV-database and DEFUSE topics specifically, no fully “neutral” primary source exists in the accessible public record at all — China has not released the database contents or confirmed/denied that DEFUSE-adjacent experiments were performed via another route, so this sub-area’s evidence base is inherently asymmetric (leaked/circumstantial documents on the lab-leak side, near-silence on the zoonosis side beyond “no confirming evidence has surfaced”) — noted per this run’s absence-of-evidence caveat rather than treated as itself probative of either hypothesis.

Graph View