Datasetpublic

depinwang/jinyang-omentum-pds-consensus-clustering-results-v1

jinyang-omentum-pds-consensus-clustering-results-v1 Null-result fallback report for the jinyang-omentum-pds-consensus-clustering canary (N=106 PDS-only real data run). Real (unpermuted) pipeline hit the pre-specified '<20 BH-FDR candidates' stopping gate (5 candidates found) -- a valid, reportable outcome, not a pipeline bug. Per design, clustering/KM/permutation-null artifacts were not produced in this branch. This is effectively the TERMINAL result for the real-data question:… See the full description on the dataset page: https://huggingface.co/datasets/depinwang/jinyang-omentum-pds-consensus-clustering-results-v1.

sourceHugging Facemitupdated 11d agoView on Hugging Face

0likes72downloads

Dataset Card

jinyang-omentum-pds-consensus-clustering-results-v1

Null-result fallback report for the jinyang-omentum-pds-consensus-clustering canary (N=106 PDS-only real data run). Real (unpermuted) pipeline hit the pre-specified '<20 BH-FDR candidates' stopping gate (5 candidates found) -- a valid, reportable outcome, not a pipeline bug. Per design, clustering/KM/permutation-null artifacts were not produced in this branch. This is effectively the TERMINAL result for the real-data question: since the Cox-selection gate itself failed (not a downstream clustering/stability gate), a 'final' mode (N=200 permutations) rerun would reproduce the byte-identical real-data outcome and immediately hit the same gate -- no further compute is needed to answer this experiment's primary question.

Dataset Info

—Rows: 1
—Columns: 12

Columns

Column	Type	Description
outcome	Value('large_string')	GATE_FAILURE
gatefailedat	Value('large_string')	Which gate failed: lt20_candidates
interpretation	Value('large_string')	Human-readable interpretation of the outcome (branched on gatefailedat)
ncandidatesbh_fdr	Value('int64')	Number of events with BH-FDR<0.05 (5, need >=20)
nexclnonconv	Value('int64')	Count of events excluded from the BH-FDR denominator due to non-convergent Cox fits
fracexclnonconv	Value('float64')	nexclnonconv / 17151
effnlow	Value('int64')	Count of events with effective n < 80 (defensive floor, expected non-binding)
effdeathslow	Value('int64')	Count of events with effective deaths < 35 (defensive floor)
confound_note	Value('large_string')	Why the RNA-quality confound check is NOT_RUN for this cohort
eventfunneljson	Value('large_string')	JSON: 24319 total -> 17151 high-confidence, by type
candidateeventsjson	Value('large_string')	JSON: the 5 BH-FDR<0.05 events with full Cox stats
reproducibility_json	Value('large_string')	JSON: R version, ConsensusClusterPlus/survival package versions, RNG seed

Generation Parameters

json

{
  "script_name": "01_pipeline.R",
  "model": "n/a (statistical pipeline: coxph + ConsensusClusterPlus, R 4.4.3)",
  "experiment_name": "jinyang-omentum-pds-consensus-clustering",
  "job_id": "turso:75360147",
  "cluster": "turso",
  "artifact_status": "partial",
  "canary": true,
  "description": "Null-result fallback report for the jinyang-omentum-pds-consensus-clustering canary (N=106 PDS-only real data run). Real (unpermuted) pipeline hit the pre-specified '<20 BH-FDR candidates' stopping gate (5 candidates found) -- a valid, reportable outcome, not a pipeline bug. Per design, clustering/KM/permutation-null artifacts were not produced in this branch. This is effectively the TERMINAL result for the real-data question: since the Cox-selection gate itself failed (not a downstream clustering/stability gate), a 'final' mode (N=200 permutations) rerun would reproduce the byte-identical real-data outcome and immediately hit the same gate -- no further compute is needed to answer this experiment's primary question.",
  "hyperparameters": {
    "canary_n_permutations_planned": 50,
    "permutations_run": 0,
    "reason_permutations_skipped": "real data gate failure -- per design, the permutation-null loop is skipped when the real pipeline itself fails the Cox-selection gate (there is nothing to circularity-correct)"
  },
  "input_datasets": []
}

Usage

python

from datasets import load_dataset

dataset = load_dataset("depinwang/jinyang-omentum-pds-consensus-clustering-results-v1", split="train")
print(f"Loaded {len(dataset)} rows")