Peer-reviewed version of S-15 - China CDC Research Square preprint — Huanan market environmental & animal surveillance (Nature 631:402–408). Confirms 74/923 environmental samples positive (70 RT-qPCR + 4 by Bowtie2 read mapping), 0/457 animal-source samples positive, ~87.5% of positives concentrated in the market’s west wing (wildlife stalls), and recovers 4 near-complete SARS-CoV-2 genomes from environmental swabs (labelled A20 = lineage A; B5, F13, F54 = lineage B), both found at the market. This is the paper whose raw sequencing reads (not originally released alongside it) were later uploaded to GISAID/NGDC in Mar 2023, triggering the Crits-Christoph/Débarre reanalyses. relevance_note: the canonical peer-reviewed China CDC market-environmental dataset; nearly every other node in this slice reanalyzes or critiques this one.
Results — sampling
O-16 - 74 of 923 Huanan market environmental samples tested positive for SARS-CoV-2
Sampling began only after the market was closed on 1 Jan 2020. The near-perfect identity of the isolated market viruses to the human epidemic virus confirms the same virus contaminated the market environment.
Link to original
O-17 - None of 457 animal-source samples from the Huanan market tested positive for SARS-CoV-2
The animal samples were collected weeks after the 1 Jan closure and removal of live animals, so they could not include the live susceptible wildlife that had been on sale; see the argument that this negative is uninformative rather than exculpatory.
Link to original
A-31 - The animal-negative result is uninformative because sampling began after animals were removed
Absence of evidence is evidence of absence only when the sampling could have detected the evidence had it been present. Here the live raccoon dogs, civets and other susceptible mammals that a wildlife-spillover scenario implicates had already been cleared out before any animal swab was taken; the 457 samples were unsold frozen carcasses, strays and fish-tank water. A negative result from a population that structurally excludes the relevant animals cannot lower the probability of animal infection at the market, so this observation should not be treated as counter-evidence to wildlife emergence.
Validity assessment — approved (checked)
Reconstruction. P1: animal sampling began 18 Jan 2020. P2: the live susceptible mammals were removed at/before the 1 Jan market closure. P3: the 457 samples were non-target material (frozen carcasses, strays, fish-tank water). Load-bearing step: a negative result drawn from a population that structurally excludes the target animals cannot lower P(animal was infected) — i.e. this is exactly the case where “absence of evidence” is not “evidence of absence,” because the detection precondition (sampling the relevant population when it could still test positive) fails.
Evaluation. Conditional on the premises the step is a correct application of the likelihood condition: for a negative to be evidence against animal infection you need P(negative | animals infected) < P(negative | animals not infected), which requires the sampling to have been able to catch an infected target animal. If the targets were gone, that inequality collapses toward equality and the update is ~null. Probed undercutting defeater: residual viral RNA on frozen product from the same supply chain could carry some weight even after live animals were cleared — but the statement claims only “almost no” weight, which already accommodates this residual channel, so the defeater does not break the link. How much weight the 0/457 actually deserves (vs. exactly zero) is a step-8 magnitude question; the inferential direction holds.
Link to original
Results — spatial and genomic
O-18 - SARS-CoV-2-positive environmental samples were concentrated in the market's western (wildlife) wing
The positive swabs are not uniformly distributed across the market but cluster in the western/wildlife section, the stalls where live mammals were sold.
Link to original
O-19 - Both early lineages A and B recovered from environmental swabs in the market's west wing
Recovery of both lineages from the market environment matters because a market-origin scenario for both lineages predicts both should be detectable there; the four genomes were all from the west-wing wildlife area.
Link to original
O-20 - Positive environmental samples contained abundant genetic material from wildlife species
The metagenomic reads place substantial wildlife biological material in the same market section as the viral contamination; the fine-grained per-swab co-location of specific susceptible species with viral reads was the object of later external reanalyses.
Link to original
A-32 - Spatial clustering of virus in the wildlife wing co-located with wildlife material fits wildlife-market emergence
If SARS-CoV-2 entered via infected live mammals held at specific wildlife stalls, environmental contamination should (i) concentrate spatially where those animals were kept, (ii) co-occur on swabs with the biological material of those species, and (iii) include both circulating lineages if both were being sold. All three hold in the dataset. This makes the data more likely under wildlife-market emergence than under a diffuse or non-market source. Caveat for validity: the same west-wing spatial signal could also be produced by an infected human working or shedding in the wildlife section, so co-location is strongly suggestive but not by itself proof that an animal, rather than a human, was the shedding source.
Validity assessment — approved (checked)
Reconstruction. The conclusion is the evidential one — “raising the probability of wildlife-associated market emergence” — not a proof. Load-bearing step: the three co-occurring features (west-wing spatial concentration, co-occurrence with wildlife-species genetic material, both lineages present) are jointly more expected under “infected wildlife at those stalls seeded the outbreak” than under a diffuse/non-market source, hence P(data | wildlife emergence) > P(data | alternative) and the posterior on wildlife emergence rises.
Evaluation. The chief undercutting defeater — an infected human working/shedding in the wildlife section reproduces the same west-wing signal — is the one the argument itself flags. Crucially that defeater only blocks a proof claim; it does not overturn the raises-probability claim, because the co-location with abundant wildlife-species material is more probable under an animal source (expected) than under a human source (coincidental), so the likelihood ratio still exceeds 1. The statement is already stated in the surviving weaker form (“raising the probability”), so no correction is needed. The strength of the update, and whether the wildlife-material co-location is as discriminating as claimed, are priced downstream; the inferential direction holds.
Link to original
Discussion
H-14 - The market environmental pattern reflects wildlife-associated emergence at the Huanan market
The wildlife-market-emergence reading of the environmental dataset. The China CDC authors themselves state the data only that the market was a site of early SARS-CoV-2 presence and are cautious about asserting origin.
Link to original
H-15 - The market may have been an amplification venue seeded by infected humans, not necessarily the origin
The cautious/agnostic reading the China CDC authors leave open: environmental swabs record where virus was shed, not which species shed it, so an infected-human-seeded market is compatible with the same data.
Link to original
A-33 - Environmental swabs cannot distinguish whether humans or animals deposited the virus
A positive environmental swab is a record of virus shed onto a surface; it is agnostic about the shedder. Infected humans (vendors, customers, early patients who frequented the market) can deposit exactly the same environmental signal as infected animals, including in the wildlife section if human activity concentrates there. Therefore the environmental positivity and its spatial pattern, taken alone, are consistent both with the market being the spillover origin and with its being a downstream amplification site seeded by infected people. This is why the China CDC authors describe the origins as remaining uncertain; the observation constrains where early virus was present but not the direction of the human-animal transfer.
Validity assessment — approved (checked)
Reconstruction. P1: a positive swab records where RNA was deposited, not which species deposited it. P2: infected humans can deposit the same signal (including a west-wing concentration if human activity concentrates there). Load-bearing step: therefore the environmental positivity and its spatial pattern, taken by themselves, are consistent with both origin (animal spillover at the market) and amplifier (humans seeded a downstream site), so these data alone do not fix the direction of transfer.
Evaluation. This is a modest underdetermination claim, and it is the logical complement of A-32: A-32 shows the metagenomic co-location raises the probability of an animal source (LR>1) but stops short of proof; A-33 makes the matching point that the presence/location data cannot definitively fix direction. The two are consistent — “some evidence” is not “distinguishes.” Probed defeater: does the wildlife-material co-location let the swabs distinguish after all? No, because that channel only shifts a probability, it does not determine the shedder for any given positive; a human shedding among wildlife material yields the same swab. The claim as stated (“cannot by themselves distinguish … leaving the question open”) survives. Valid.
Link to original