[
 {
  "local_id": "C1",
  "type": "empirical",
  "core": true,
  "statement": "Of the 37 well-powered primary claims (testable signed headline claims of protein Mendelian randomization papers in PubMed, January 2024 to September 2026, whose FinnGen R12 or Pan-UK Biobank outcome study matches the disease and is expected to reach |z| of at least 5 at the protein's lead for the claim's own odds ratio per SD, at most 1.2, or 1.1 where it gives none), 10 (27.0%; exact 95% CI 13.8% to 44.1%) hold up in the pre-registered re-test: their Wald ratio at a UK Biobank Pharma Proteomics Project cis-pQTL has the claimed sign with p < 0.05 and a cis credible set colocalises with the disease (PP.H4 at least 0.8).",
  "evidence": [
   {
    "result": "R1.hold_up_of",
    "produced_by": "code/run"
   },
   {
    "result": "R1.hold_up",
    "produced_by": "code/run"
   },
   {
    "result": "R1.hold_up_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R1.hold_up_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R1.hold_up_high_pct",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives another count of well-powered primary claims that hold up.",
  "confidence": 0.8
 },
 {
  "local_id": "C2",
  "type": "empirical",
  "core": true,
  "statement": "On the measure without a direction, colocalisation with p < 0.05 either way, the 37 well-powered primary claims score 32.4% against 0.9% for three random UK Biobank Pharma Proteomics Project proteins each, drawn from those equally well powered in the same outcome study and claimed by no paper for the disease, a difference of 31.5 percentage points (95% CI 16.2 to 47.7, from 10,000 resamples of claims).",
  "evidence": [
   {
    "result": "R2.claims",
    "produced_by": "code/run"
   },
   {
    "result": "R2.claims_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R2.controls_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R2.difference_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R2.low_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R2.high_pp",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives a difference whose 95% interval includes zero.",
  "confidence": 0.85
 },
 {
  "local_id": "C3",
  "type": "empirical",
  "core": true,
  "statement": "On the same measure, all 339 primary claims score 10.9% against 0.2% for three random proteins each from the same fifth of instrument strength, tested in the same outcome study, a difference of 10.7 percentage points (95% CI 7.5 to 14.2).",
  "evidence": [
   {
    "result": "R3.claims",
    "produced_by": "code/run"
   },
   {
    "result": "R3.claims_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R3.controls_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R3.difference_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R3.low_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R3.high_pp",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives a difference whose 95% interval includes zero.",
  "confidence": 0.85
 },
 {
  "local_id": "C4",
  "type": "empirical",
  "core": true,
  "statement": "Of all 339 primary claims, 31 (9.1%; 95% CI 6.3% to 12.7%) hold up, 79 (23.3%) fail, and 229 (67.6%; 95% CI 62.3% to 72.5%) are inconclusive, and of the 110 the outcome data judged either way, 31 (28.2%; 95% CI 20.0% to 37.6%) hold up.",
  "evidence": [
   {
    "result": "R4.holds_up_of",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.fail",
    "produced_by": "code/run"
   },
   {
    "result": "R4.fail_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.inconclusive",
    "produced_by": "code/run"
   },
   {
    "result": "R4.inconclusive_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.inconclusive_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.inconclusive_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_when_judged_of",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_when_judged",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_when_judged_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_when_judged_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R4.holds_up_when_judged_high_pct",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives other counts of primary claims that hold up, fail, or are inconclusive.",
  "confidence": 0.85
 },
 {
  "local_id": "C5",
  "type": "empirical",
  "core": false,
  "statement": "Of the 37 well-powered primary claims, 16 reproduce (43.2%), 12 colocalise (32.4%), 13 fail (35.1%), 14 are inconclusive (37.8%), and none has a better-supported neighbouring gene (95% CI 0% to 9.5%).",
  "evidence": [
   {
    "result": "R5.reproduce_of",
    "produced_by": "code/run"
   },
   {
    "result": "R5.reproduce",
    "produced_by": "code/run"
   },
   {
    "result": "R5.reproduce_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R5.colocalise",
    "produced_by": "code/run"
   },
   {
    "result": "R5.colocalise_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R5.fail",
    "produced_by": "code/run"
   },
   {
    "result": "R5.fail_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R5.inconclusive",
    "produced_by": "code/run"
   },
   {
    "result": "R5.inconclusive_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R5.better_supported_neighbour",
    "produced_by": "code/run"
   },
   {
    "result": "R5.better_supported_neighbour_high_pct",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [
   "C1"
  ],
  "falsified_if": "Re-running code/run gives other counts for these measures.",
  "confidence": 0.8
 },
 {
  "local_id": "C6",
  "type": "empirical",
  "core": false,
  "statement": "Among the well-powered primary claims, 0 of 5 tested in outcome data independent of the original's hold up, against 4 of 18 in overlapping data and 6 of 14 where the overlap is unknown.",
  "evidence": [
   {
    "result": "R6.independent",
    "produced_by": "code/run"
   },
   {
    "result": "R6.independent_of",
    "produced_by": "code/run"
   },
   {
    "result": "R6.overlapping",
    "produced_by": "code/run"
   },
   {
    "result": "R6.overlapping_of",
    "produced_by": "code/run"
   },
   {
    "result": "R6.unknown",
    "produced_by": "code/run"
   },
   {
    "result": "R6.unknown_of",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [
   "C1"
  ],
  "falsified_if": "Re-running code/run gives other counts by independence.",
  "confidence": 0.75
 },
 {
  "local_id": "C7",
  "type": "empirical",
  "core": false,
  "statement": "Counting claims of one protein, outcome study, and direction once, 10 of 36 well-powered primary claims hold up (27.8%; 95% CI 14.2% to 45.2%).",
  "evidence": [
   {
    "result": "R7.hold_up",
    "produced_by": "code/run"
   },
   {
    "result": "R7.hold_up_of",
    "produced_by": "code/run"
   },
   {
    "result": "R7.hold_up_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R7.hold_up_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R7.hold_up_high_pct",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [
   "C1"
  ],
  "falsified_if": "Re-running code/run gives another count.",
  "confidence": 0.8
 },
 {
  "local_id": "C8",
  "type": "empirical",
  "core": false,
  "statement": "The share of the primary claims that hold up rises with the outcome's expected |z| at the rule's effect, from 4.0% (8 of 198) below 2 and 8.5% (7 of 82) from 2 to 4 to 27.3% (6 of 22) from 4 to 5, 25.0% (6 of 24) from 5 to 8, and 30.8% (4 of 13) at 8 or more.",
  "evidence": [
   {
    "result": "R8.below_2_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R8.below_2",
    "produced_by": "code/run"
   },
   {
    "result": "R8.below_2_of",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_2_to_4_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_2_to_4",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_2_to_4_of",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_4_to_5_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_4_to_5",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_4_to_5_of",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_5_to_8_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_5_to_8",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_5_to_8_of",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_8_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_8",
    "produced_by": "code/run"
   },
   {
    "result": "R8.from_8_of",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives other counts by band.",
  "confidence": 0.8
 },
 {
  "local_id": "C9",
  "type": "empirical",
  "core": false,
  "statement": "With a claim that gives no effect per SD tested at an odds ratio of 1.2 instead of 1.1, 14 of 73 well-powered primary claims hold up (19.2%; 95% CI 10.9% to 30.1%), and with it tested at the median effect that claims of similar power report, 10 of 32 (31.2%; 95% CI 16.1% to 50.0%).",
  "evidence": [
   {
    "result": "R9.default_1_2",
    "produced_by": "code/run"
   },
   {
    "result": "R9.default_1_2_of",
    "produced_by": "code/run"
   },
   {
    "result": "R9.default_1_2_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R9.default_1_2_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R9.default_1_2_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R9.similar_power",
    "produced_by": "code/run"
   },
   {
    "result": "R9.similar_power_of",
    "produced_by": "code/run"
   },
   {
    "result": "R9.similar_power_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R9.similar_power_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R9.similar_power_high_pct",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [
   "C1"
  ],
  "falsified_if": "Re-running code/run gives other counts for either variant of the rule.",
  "confidence": 0.8
 },
 {
  "local_id": "C10",
  "type": "empirical",
  "core": false,
  "statement": "Of the 223 unsigned claims whose outcome study matches the disease and whose estimate doesn't rest only on non-European data, 40 (17.9%; 95% CI 13.1% to 23.6%) colocalise with p < 0.05, against 0.3% of their controls, a difference of 17.6 percentage points (95% CI 12.7 to 23.0).",
  "evidence": [
   {
    "result": "R10.nodir_of",
    "produced_by": "code/run"
   },
   {
    "result": "R10.nodir",
    "produced_by": "code/run"
   },
   {
    "result": "R10.nodir_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R10.nodir_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R10.nodir_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R10.controls_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R10.difference_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R10.low_pp",
    "produced_by": "code/run"
   },
   {
    "result": "R10.high_pp",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run gives a difference whose 95% interval includes zero.",
  "confidence": 0.85
 },
 {
  "local_id": "C11",
  "type": "negative_result",
  "core": false,
  "statement": "The dose-response test failed its pre-registered calibration on Open Targets' curated gold-standard loci: 8 of the 76 judged loci could take it, fewer than the 30 the gate required, its sensitivity was 37.5% (95% CI 0.0% to 71.4%), and the 1 decoy that could take it passed, so the test serves as description only.",
  "evidence": [
   {
    "result": "R11.testable_gold_loci",
    "produced_by": "code/run"
   },
   {
    "result": "R11.judged_loci",
    "produced_by": "code/run"
   },
   {
    "result": "R11.sensitivity_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R11.sensitivity_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R11.sensitivity_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R11.decoys",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run on the judged loci gives at least 30 testable loci, or a gate that passes.",
  "confidence": 0.9
 },
 {
  "local_id": "C12",
  "type": "empirical",
  "core": false,
  "statement": "Among the 37 well-powered primary claims, the dose-response test, which failed its calibration, passes for 8, fails for 21, and can't run for 8.",
  "evidence": [
   {
    "result": "R12.passes_of",
    "produced_by": "code/run"
   },
   {
    "result": "R12.passes",
    "produced_by": "code/run"
   },
   {
    "result": "R12.fails",
    "produced_by": "code/run"
   },
   {
    "result": "R12.cannot_run",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [
   "C1",
   "C11"
  ],
  "falsified_if": "Re-running code/run gives other counts.",
  "confidence": 0.85
 },
 {
  "local_id": "C13",
  "type": "empirical",
  "core": false,
  "statement": "Of the 2,432 PubMed records of January 2024 to September 2026 matching a query for protein Mendelian randomization, 989 (40.7%) make an eligible claim of a circulating protein's effect on a disease, and 847 (34.8%) are excluded because their exposure isn't a protein's circulating level, among them 386 using gene expression or other molecular QTLs, 219 with no protein exposure, and 177 weighting a drug target by a biomarker.",
  "evidence": [
   {
    "result": "R13.papers",
    "produced_by": "code/run"
   },
   {
    "result": "R13.eligible",
    "produced_by": "code/run"
   },
   {
    "result": "R13.eligible_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R13.e2",
    "produced_by": "code/run"
   },
   {
    "result": "R13.e2_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R13.e2_expression",
    "produced_by": "code/run"
   },
   {
    "result": "R13.e2_no_protein",
    "produced_by": "code/run"
   },
   {
    "result": "R13.e2_biomarker",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run on the coded census gives other counts.",
  "confidence": 0.85
 },
 {
  "local_id": "C14",
  "type": "empirical",
  "core": false,
  "statement": "Of the 989 papers with an eligible claim, 638 (64.5%; 95% CI 61.4% to 67.5%) have one testable with public data, while for 209 the UK Biobank Pharma Proteomics Project doesn't measure the protein and for 136 it has no cis signal.",
  "evidence": [
   {
    "result": "R14.testable_of",
    "produced_by": "code/run"
   },
   {
    "result": "R14.testable",
    "produced_by": "code/run"
   },
   {
    "result": "R14.testable_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R14.testable_low_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R14.testable_high_pct",
    "produced_by": "code/run"
   },
   {
    "result": "R14.not_measured",
    "produced_by": "code/run"
   },
   {
    "result": "R14.no_cis",
    "produced_by": "code/run"
   }
  ],
  "depends_on": [],
  "falsified_if": "Re-running code/run on the coded census gives other counts.",
  "confidence": 0.9
 }
]
