Oncofusion

start / graph / absence-the-supplementary-tables-behind-three-corpus-claims-cannot-be-retrieved-from-pmc

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

Bounded unreachable: the Blood supplemental table that three entries read their NLRC5 and CIITA group assignments from cannot be retrieved by ANY programmatic route - PMC serves a landing page and the publisher 403s this client - so those claims are re-readable but not re-verifiable without a human's institutional access

also called supplemental Table 4 unreachable · PMC serves a landing page instead of the xlsx · Zhou 9,292-locus table not re-fetchable · 22 supplementary files, 0 retrieved · publisher route blocked 403, not absent

Three entries in this graph rest on numbers read out of a spreadsheet attached to a published paper. This entry exists so a future cycle does not waste time trying to download that spreadsheet: every route tried returns either a missing-file error or a web page pretending to be the file.

The same thing, for a scientist

🟠 **UPDATED 2026-09-24 (cycle 135): the publisher route named below as *untried, not exhausted* has now been tried. It is BLOCKED, which is a third outcome and not the same as absent.** | route | result | |---|---| | `doi.org` HTML redirect | **HTTP 403** | | `ashpublications.org` article landing page | **HTTP 403** | | `ashpublications.org` main PDF (the authoritative link Crossref registers) | **HTTP 403** | | **`ashpublications.org/` host root** | **HTTP 403** | | Crossref metadata for the DOI | **HTTP 200 - and it registers NO supplement link**, only the main PDF, twice | **The host refuses this client for everything, including its own front page**, so no guessed supplement path could have succeeded and none was tried. This is a **client/network block, not a missing file**: the supplement almost certainly exists and is served to a browser with institutional access. 🔵 **So the gap is now precisely located, and it is not a research gap.** Every programmatic route is exhausted - PMC serves a landing page, the PMC OA service 404s, and the publisher 403s this client. **A human with Blood access can open `BLOOD_BLD-2023-023644-mmc4.xlsx` in one click and settle the three claims.** That is the cheapest remaining action on this thread and it is not one an agent can take. --- **This entry exists so that a future cycle does not repeat the search.** Three entries - `ana-the-transactivator-split-is-two-genes-and-the-mhc-is-invisible`, `gene-nlrc5` and `gene-ciita` - attribute specific group assignments to **Zhou's published 9,292-locus menin and H2AK119ub target set**, with genomic coordinates (`NLRC5` menin-solo at `chr16:56989555`; `CIITA` H2AK119ub-solo at `chr16:10877200`). The source is **supplemental Table 4** of PMID **39158067** (Zhou et al., *Blood* 2024, DOI 10.1182/blood.2023023644). **The full text of that paper does not contain the strings `NLRC5` or `CIITA` anywhere**, which is expected - the assignments live in the supplement, and PMC's full-text XML does not carry supplementary file contents. **Every route tried to reach the supplement failed:** | route | result | |---|---| | PMC `/bin/BLOOD_BLD-2023-023644-mmc4.xlsx` | **HTTP 404** | | PMC `/bin/` for mmc1, mmc2, mmc3 | **HTTP 200 serving `text/html`**, ~21,400 bytes, body begins `<!do` - a landing page, not the file | | PMC Open Access service (`oa.fcgi?id=PMC11561541`) | **HTTP 404** | Across all six papers examined this cycle the pattern is uniform: **22 supplementary files listed, 0 retrieved** - eight HTTP 200 landing pages and the rest 404 [E1]. 🟠 **What this does and does not license.** It does **not** cast doubt on the three claims. The citing entries record that the table was *read directly* at the time, and `ana-the-transactivator-split...` records independent confirmation by a separate beat. What it establishes is narrower and still worth writing down: **those claims are currently re-readable but not re-verifiable**, and any future audit that treats "the term is absent from the full text" as evidence against them will be wrong for the same reason this cycle nearly was. 🔵 ~~**The route not tried, named so it is not mistaken for exhausted:** the publisher-hosted supplement behind DOI 10.1182/blood.2023023644 on ashpublications.org.~~ **TRIED IN CYCLE 135 - see the banner at the top: the host returns 403 to this client for every URL including its own root.** This entry bounds what **any programmatic route** will serve, not what exists.

Go deeper
Why a 200 here is worse than a 404

Eight of the twenty-two supplement URLs return HTTP 200. A naive fetcher records success, stores ~21 KB, and - if it is scoring text - finds none of the expected terms in what is actually a navigation page. The result is a confident negative about a citation, generated entirely by the archive's error handling.

The tell is cheap and should be standard: the bodies are text/html, they begin <!do, and a .pptx and a .pdf came back four bytes apart in size. A supplementary file is PK.. (zip-based Office formats) or %PDF. Require the magic bytes, not the status code.

This is the same failure shape this corpus already records for HTTP 429 read as absence, and for an HTTP 200 without an <article> element.

Facts and figures
target paperPMID 39158067 / PMC11561541 / DOI 10.1182/blood.2023023644
target supplementBLOOD_BLD-2023-023644-mmc4.xlsx (captioned 'Supplemental Table 4')
route not triednone remaining programmatically; requires a human with Blood institutional access to open BLOOD_BLD-2023-023644-mmc4.xlsx
supplementary files seen22
supplementary files retrieved0
crossref registers supplement linkfalse
blocked is not absentthe 403 is a property of this client and network; the file is almost certainly served to an authenticated browser
absence classmeasured-not-curated
absence class basisThe tables were measured AND published; every retrieval route is blocked (doi.org 403, publisher landing page). This entry's own words: BLOCKED is a third outcome and not the same as absent - which the controlled vocabulary cannot express, so the nearest token is used and this note carries the truth.
entries depending on it
[
  "ana-the-transactivator-split-is-two-genes-and-the-mhc-is-invisible",
  "gene-nlrc5",
  "gene-ciita"
]
routes tried
{
  "pmc_bin_path": "404 for mmc4; HTTP 200 text/html landing page for mmc1-3",
  "pmc_oa_service": "404",
  "europepmc_fulltextxml": "503/500 all day; NCBI efetch used instead and succeeded",
  "publisher_ashpublications": "HTTP 403 on the article landing page, on the authoritative main-PDF link and on the host root - a client/network block, not a missing file",
  "doi_org_html": "HTTP 403"
}
fulltext contains
{
  "menin": true,
  "H2AK119ub": true,
  "NLRC5": false,
  "CIITA": false
}
measurements
[
  {
    "quantity": "supplementary files retrieved as actual files",
    "value": 0,
    "unit": "count",
    "subject": "the first three supplementary files of each of 6 papers, probed via the PMC /bin/ path",
    "universe": "22 supplementary files listed in the PMC XML of those 6 papers"
  }
]
Still open
  1. CLOSED programmatically in cycle 135: ashpublications.org returns 403 to this client on the article page, the Crossref-registered PDF and the host root, and Crossref registers no supplement link. What remains is not an agent task - a human with Blood institutional access can open BLOOD_BLD-2023-023644-mmc4.xlsx directly and settle the NLRC5/CIITA assignments in one click.
  2. Is the HTTP 200 landing page behaviour specific to these publishers, or does PMC do it generally? If general, every supplement-derived claim in this corpus is in the same position and the count is much larger than three.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

    Across the six papers behind the lowest-scoring claim/citation pairs, 22 supplementary files are listed in the PMC XML and none could be retrieved: the probed URLs return either HTTP 404 or HTTP 200 with content-type text/html and a body of roughly 21,400 bytes beginning with a doctype, and the PMC Open Access service returns 404 for PMC11561541.

    What this evidence cannot show

    Run 2026-09-24 (cycle 133). Only the first three supplementary files per paper were probed, so the census is complete for 14 of the 22 and inferred for the remainder. Two routes were tried, both hosted by PMC. The publisher's own copy behind the DOI was not tried. A retry on two URLs that initially returned 200 produced 404, so the behaviour is not even stable within a single session.

    Evidence that argues against this (1)
    • Unreachable now is not unreachable then, and certainly not absent. The citing entries were written when the table was read, and one records confirmation by a second, independent beat. A future cycle that reaches the publisher-hosted copy would supersede this entry's practical value entirely, which is why the untried route is named in the summary rather than buried.
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.

    The full text of PMID 39158067 contains the strings 'menin' and 'H2AK119ub' but does not contain 'NLRC5' or 'CIITA' anywhere, while its body does reference 'supplemental Table 4' and the article lists seven supplementary files including one captioned 'Supplemental Table 4'.

    PMID 39158067DOI
    What this evidence cannot show

    Full text fetched 2026-09-24 via NCBI efetch db=pmc for PMC11561541 (119 KB of XML) and searched as plain text after tag stripping; EuropePMC was returning HTTP 503/500 for all identifiers that day. String search is case-insensitive and word-bounded, so a hyphenated or superscripted rendering of a symbol could in principle be missed.

    Evidence that argues against this (1)
    • The absence of the two symbols from the full text is exactly what the citing entries predict, since they attribute the assignments to a supplemental table rather than to the body. This evidence item therefore corroborates the entries' own account of where their data came from, and reading it the other way - as the paper failing to mention the genes - would be the error this entry is written to prevent.
Connected to