← graph

fact-check-validate

Fact-check: validate all scientific claims in research documents

Metadata

Statusdone
Assignedagent-540
Agent identity3184716484e6f0ea08bb13539daf07686ee79d440505f1fdf2de0357707034c3
Created2026-04-02T14:25:05.235513239+00:00
Started2026-04-02T14:25:24.667536654+00:00
Completed2026-04-02T14:31:14.182483723+00:00
Tagsvalidation,critical, eval-scheduled
Eval score0.82
└ blocking impact0.85
└ completeness0.95
└ coordination overhead0.88
└ correctness0.85
└ downstream usability0.90
└ efficiency0.75
└ intent fidelity0.78
└ style adherence0.85

Description

CRITICAL CONTEXT

PHR = Pseudohomologous Region. Read subtelomeric_analysis_report.md and TODO.md for authoritative project context.

Goal

Systematically fact-check every scientific claim in the deep research documents. These were written by LLM agents that are known to confabulate — the PHR terminology error ('Palindromic Heterochromatin Repeats') proves this. We need to identify EVERY claim that is stated as fact and verify it has a real basis.

Documents to fact-check

  1. deep_research_olfactory_receptors.md
  2. deep_research_dux4_frg2.md
  3. deep_research_tubb8.md
  4. deep_research_gtp_binding.md
  5. deep_research_synthesis.md

For EACH document, check:

1. Literature citations

  • Are papers cited by author name and year? Are these REAL papers?
  • Flag any citation that looks suspicious (e.g., wrong author-year combo, non-existent journals)
  • Where possible, verify via DOI or PubMed ID
  • ADD real DOIs/PMIDs where they're missing but the claim is real
  • REMOVE fake citations

2. Gene function claims

  • Is each gene's described function correct? Cross-reference with NCBI Gene, UniProt, OMIM
  • DUX4: Is the FSHD mechanism correctly described? D4Z4 repeat contraction, DUX4 de-repression, etc.
  • TUBB8: Is it really oocyte-specific? Do mutations really cause female infertility?
  • OR4F: Are these really functional olfactory receptors or pseudogenes?
  • IQSEC3: Is it really a postsynaptic density protein in GABAergic synapses?
  • IL9R: Is the JAK/STAT signaling pathway correctly described?

3. Numerical claims

  • Copy numbers should match gene_copy_summary.csv
  • Enrichment p-values should match the CSV files
  • Gene counts should match our actual data

4. Evolutionary/mechanistic claims

  • Is the subtelomeric OR gene shuffling mechanism (Trask et al.) real?
  • Is the D4Z4 inter-chromosomal exchange story correctly told?
  • Are claims about population-specific enrichment (AFR, SAS) supported by the data?

5. Disease associations

  • Are OMIM numbers correct (if cited)?
  • Are disease-gene links real (DUX4-FSHD, TUBB8-infertility, IL9R-asthma, SHOX-Turner syndrome)?

Output

For each document, produce a checklist:

  • Claim | Verified? | Source/DOI | Notes

Save as fact_check_report.md

Mark each claim as:

  • ✅ VERIFIED — confirmed with real source
  • ⚠️ PLAUSIBLE — sounds right but couldn't verify specific citation
  • ❌ WRONG — incorrect or fabricated
  • 🔍 UNVERIFIABLE — can't confirm or deny

Focus on getting this RIGHT, not on being fast. If you're unsure about a claim, mark it as unverifiable rather than rubber-stamping it.

Validation

  • Every research document is covered
  • Every major claim has a verification status
  • Fake citations are identified
  • Real DOIs/PMIDs are added where possible

Depends on

Required by

Log