Testing the PX1 signal.
Keeping the unknowns visible.

A Python standard-library program inspected the released example VCF at three specified PX1 SNP positions, preserved mixed calls separately, and matched the author’s K13 metadata by sample ID. No upstream R code, generated code or wet-lab procedure ran.

VCF samples

211

Metadata matches

211

Exit code

0

Clean rerun

Identical output

COMPUTATIONAL OBSERVATION — NOT INDEPENDENTLY VALIDATED

• all three alt present unphased: 8 samples
• all three pure alt: 106 samples
• other complete: 97 samples

These counts describe an author-selected example. They are neither PIN haplotype prevalence nor an estimate of resistance prevalence. The defining deletions, phase and phenotype are not established by this inspection.

1. 2026-10-08T17:53:59.552085+00:00 · UTC

   Formal public-artifact inspection started with the pinned author release and recorded inputs.

2. 2026-10-08T17:54:07.039042+00:00 · UTC

   Inspection completed, exit 0, 7487 ms. The central model remained blocked at input validation.

3. Separate process reruns returned byte-identical output. Their scope was artifact inspection, not the phenotype model.
4. Next: recover model inputs and independently verify haplotype definitions before extending to other African datasets.

The scientific interpretation remains attached to the primary record and its supplied artifacts.

• Primary paper · DOI 10.1038/s41591-026-04590-5
• Author’s archived code and example data · 10.5281/zenodo.21283869
• Pinned author repository · MIT license

Source check recorded 2026-10-15T22:08:28.283Z (UTC). This is a dated input catalogue, not a live availability guarantee or another experiment result.

• Main-paper source data located. The publisher-linked source workbook has sample names, haplotypes, drug-response fields and parasitemia. Of the 211 example sample names, 161 match exact text keys; one numeric sample key remains unmapped. Its 1,347 physical rows include a header, not a qualified analysis denominator. COI is not a named column in this sheet; wsaf has not been substituted for it. Data-specific reuse permission and the complete model cohort remain unverified. Article terms: CC-BY-NC-ND 4.0.
• Earlier phenotype source located. The published source workbook contains seven sheets, including a data dictionary and drug-response tables. Its UID columns and parasitemia percentage unit are inspectable. The per-isolate join to this PX1 analysis is not yet established. Source paper and reuse terms.
• Pinned author inputs checked. The exact repository tree contains the model script and the 211-row example metadata. The three named model-input CSVs are absent from that tree; this does not prove they are unavailable elsewhere. The script was not executed by this source check.
• PRJNA1298911 (WGS). 158 deposited runs, 158 distinct sample accessions; 465,882,029,469 known FASTQ bytes. Every run has a listed FASTQ. Inspect provider manifest. Raw reads have not been downloaded.
• PRJNA1450510 (AMPLICON). 1,375 deposited runs, 1,375 distinct sample accessions; 6,124,279,733 known FASTQ bytes. 71 runs have no listed FASTQ; the byte total is incomplete. 546 records have no species label in this report. Inspect provider manifest. Raw reads have not been downloaded.
• Pf7 comparison metadata checked. 20,864 distinct samples across 33 known countries. Known collection years: 1984–2018; 160 rows have no parsed year. The provider’s QC-pass field is true for 16,203 rows, not proof of suitability for this analysis. Its MD5 matched the provider checksum. Genotypes, PIN calls and a compatible cohort have not been established. Pf7 release and licence.

Observed: the three target SNP records occur in the supplied example VCF, with 211 samples matched to the supplied metadata. Inspect the capsule for calls, hashes and methods.

Not established: a complete PIN haplotype, drug-response effect, independent relationship with K13, clinical treatment failure, or geographic spread.

Critic verdict: REJECT SCIENTIFIC GENERALIZATION. Selected example genotypes cannot establish drug resistance, an independent PX1 effect, or regional prevalence. Missing phenotype/covariate inputs prevent the central model reproduction.

No new scientific collision or Ghost Node has been validated from this attempt.

1. Computational: validate the full haplotype including deletions and mixed-infection handling, then rerun the author’s specified model in a pinned environment once exact phenotype/covariate inputs are recovered. This addresses a prerequisite before extending the signal geographically.
2. Additional data: locate the missing per-isolate model inputs and data dictionary, resolve cohort inclusion criteria, and establish access/usage conditions for public African genomic cohorts and resistance metadata.
3. Wet-lab plan: after computational review, design an approved isogenic genotype comparison that controls K13 background and measures assay-specific drug response. This is a proposal requiring domain expertise and institutional review; Cathedral will not actuate it.

These priorities are an artifact-availability assessment, not estimated information-gain scores. Costs, runtime, feasibility and biological discriminators remain unmeasured.

• The VCF is an author-selected example, not a representative African cohort.
• Three SNPs do not establish the full PIN haplotype; the two defining deletions and infection-level phase are not assessed.
• No genotype quality or depth threshold was applied beyond retaining the source VCF calls; mixed and missing calls are separate categories.
• K13 labels are author-supplied metadata, not independently genotyped in this analysis.
• No phenotype data, adjusted effect size, clinical outcome or causal interpretation was generated.
• No comparison to MalariaGEN or other African public cohorts was performed.

{
  "provider": "local-python",
  "tool": "public-artifact-probe",
  "version": "001-public-artifact-probe-v1",
  "startedAt": "2026-10-08T17:53:59.552085+00:00",
  "completedAt": "2026-10-08T17:54:07.039042+00:00",
  "durationMs": 7487,
  "exitCode": 0,
  "command": [
    "python",
    "challenges/001/code/probe_public_artifacts.py",
    "--vcf",
    "/workspace/scratch/feaa98d57d36/challenge-work/uganda.vcf.gz",
    "--metadata",
    "/workspace/scratch/feaa98d57d36/challenge-work/upstream/Examples (selection)/px1_genotypes_metadata_combined.tsv",
    "--upstream",
    "/workspace/scratch/feaa98d57d36/challenge-work/upstream",
    "--output",
    "/tmp/cthdrl-challenge-formal-attempt.json"
  ],
  "stdout": "{\"verdict\": \"INSUFFICIENT PUBLIC ARTIFACTS\", \"samples\": 211, \"matchedMetadata\": 211, \"missingInputs\": [\"ex_vivo_phenotype_data_long_table_current.csv\", \"parasitemia.csv\", \"COI_calls_all_uganda_samples.csv\"]}\n",
  "stderr": "",
  "outputSha256": "54363549b20ab97e6794b1d6c01a795102d735e628d01c260a80f17216f6edd3",
  "cleanRerun": {
    "executed": true,
    "identicalOutput": true,
    "scope": "Independent artifact inspection only; not a rerun of the missing phenotype model"
  },
  "repository": "https://github.com/Khufu2/cthdrl",
  "upstreamRepository": "https://github.com/Karaniare/PfWGS-DRscanner",
  "upstreamCommit": "8ef861e27f93d95859113003dc01d24747fb28c7",
  "upstreamArchiveDoi": "10.5281/zenodo.21283869",
  "upstreamLicense": "MIT",
  "paperDoi": "10.1038/s41591-026-04590-5",
  "primaryModelExecuted": false,
  "unknowns": [
    "Actual PIN haplotype including deletions",
    "Availability of additional model inputs outside the inspected release",
    "Independent K13 genotypes",
    "African cohort representativeness",
    "Phenotype effect sizes"
  ],
  "figures": [
    {
      "name": "snp-call-categories.svg",
      "sha256": "d999a0e2a22b1c551239d872073407d8f913025a76cac0e4f34f1753aa09f63d",
      "generator": "matplotlib",
      "version": "3.10.8",
      "scope": "Descriptive counts from the recorded selected example, not regional prevalence"
    }
  ]
}

This is research software and a bounded computational inspection, not patient treatment advice.