Analysis Tools · MVA · BUB1B · sensitivity

Post-hoc sensitivity audit (October 24, 2026)

Archived result
Executed reanalysis

What it is not. Post-hoc and exploratory. Diagnostic ranges are not confidence intervals; nothing here increases n or establishes biological validation.

Dates, kept apart

Archived result recorded: Versioned in the research repository; see the execution record
Executions: 2026-10-24 20:40:07 UTC · research repository
Repository commit: a434af9
This page rendered: 2026-10-25 04:58:35 UTC · a view date, not an execution or publication date

Values read from the files

Read from the versioned result files of the research repository at commit a434af9, verified by SHA-256.

Quantity	Value	Unit	Source file
PTC measured rows / distinct contexts	12 / 8	rows / contexts	sensitivity_results.json (research repository)
Exact six-base matches	0	rows	sensitivity_results.json (research repository)
Median paired increment, full set	+0.0165	percentage points	sensitivity_results.json (research repository)
Leave-one-row-out range	+0.0140 to +0.0190	percentage points (finite range, not an interval)	sensitivity_results.json (research repository)
Leave-one-context-group-out range	+0.0135 to +0.0285	percentage points (finite range, not an interval)	sensitivity_results.json (research repository)
Difference of marginal medians	+0.0240	percentage points (a different quantity)	sensitivity_results.json (research repository)
GSE65145 thresholds selecting the same features	95%, 97.5%, 99%, 99.5%	reciprocal overlap	sensitivity_results.json (research repository)
GSE65145 thresholds with no eligible feature	100%	reciprocal overlap (insufficient, not zero expression)	sensitivity_results.json (research repository)
GSE65145 Rap / Pro, arithmetic versus geometric	0.585706 versus 0.584998	ratio	sensitivity_results.json (research repository)
GSE65145 leave-one-sample-out range	0.542241 to 0.629171	ratio (finite range)	sensitivity_results.json (research repository)

Execution records

Each row is a run with a record. “Match” means the output bytes hash to the archived value: deterministic computation reproduced, nothing more.

Execution records. A matching SHA-256 is computational reproduction of a deterministic output, not biological validation.
Execution / record time	What ran	Environment	Inputs	Artifacts versus archive	Record
2026-10-24 20:40:07 UTCResearch repository	Sensitivity audit validation by the research repository (October 24, 2026)11 tests passed, 0 failed; a second run into a separate directory was byte-identical (9 files).	CPython 3.12.14 Existing environment, not isolated	14 / 14 verified by SHA-256Previously downloaded public cache inputs, each verified against unchanged original manifest; not a fresh network download.	match `robustness/results/sensitivity_results.json` SHA-256 4062d0d4d50a…	Record (JSON) At the commit

Traceability

Primary source: See the dataset and the research repository documents below.
Dataset: The same manifest-verified inputs of the PTC and GSE65145 reanalyses.
Biological model: Same models as the two reanalyses it audits.
Samples: PTC: 12 measured rows / 8 contexts (exhaustive deletions). GSE65145: 2 samples per condition, 3 redundant features.
Parameters: Protocol frozen before the calculations, after the original results were known; amendment 01 recorded before execution
Exhaustive finite-set deletions; no random seed, bootstrap, p-value or interval
Script: track2/reproduction/robustness/run_sensitivity.py · SHA-256 c12e186cb037a55324020e25dc222adea634d34d7c0502e1741502e585f1d104 · download
Command: python track2/reproduction/robustness/run_sensitivity.py; python track2/reproduction/robustness/test_sensitivity.py
Commit: ivillafuerte/mva-hackathon-2026 @ a434af99342f57aa9fde958535c173b82c315c2a

Downloads

Provenance manifest (JSON) Analysis (JSON)

The manifest lists sources, model, samples, parameters, script, commit, inputs, execution records and every file below with its SHA-256. Its generatedAt is when it was written, not an execution date.

Results

sensitivity_results.json Sensitivity results (machine-readable) · 80.4 KB · SHA-256 4062d0d4d50a2d0e6ef43904727e7f50487ae10de9dd16affc00c1ca2129c2e1 · at the commit
ptc_nearest_measured_rows.csv PTC nearest measured rows (12 rows, 8 contexts) · 865 B · SHA-256 1407778fc084d5aac4d8f7c191e588d4e6db00632347a07df66fd7f32e6b5f95 · at the commit
ptc_leave_one_variant_row_out.csv PTC leave-one-row-out deletions · 1.5 KB · SHA-256 f89938c005ad2a1ab4befea644f5c0d78738ff5cf64ddf525e396ffb44f89846 · at the commit
ptc_leave_one_context_group_out.csv PTC leave-one-context-group-out deletions · 1.1 KB · SHA-256 83370e86e2937158715a4bf8b0cdd9397b6cc888da34ce3bbef49ace1757fdba · at the commit
gse_feature_thresholds.csv GSE65145 reciprocal-overlap threshold grid · 1.6 KB · SHA-256 b434287d35b29a22fce951d398fb8a5d8a1b85dc3d175e9aa8e524de9bfcc2bb · at the commit
gse_feature_estimators.csv GSE65145 arithmetic versus geometric estimators · 1.7 KB · SHA-256 734101642d5737e38104d7252ba822836c26d931ea1e666c7b78e800d3b573e6 · at the commit
gse_leave_one_sample_out.csv GSE65145 leave-one-sample-out deletions · 6.1 KB · SHA-256 c1d00bd0937714134bd38d9451e187a3addb7fa83df8f39a7a50d5c1044b51e5 · at the commit
gse_locus_overlap_audit.csv GSE65145 locus overlap audit (18 rows) · 1.6 KB · SHA-256 5ec88abe8f47ac02a5af6ba3fbbed6dd0f50aacf734eb330f8051394fe43ccba · at the commit

Execution records

validation_summary.json Validation summary of the sensitivity audit (tests, determinism, preserved files) · 3.4 KB · SHA-256 26bee1e1ebb9e3c9c09e193e89b46b75af8a2af4b45a15015fc724da65da6aee · at the commit

Scripts

run_sensitivity.py Sensitivity analysis script · 24.7 KB · SHA-256 c12e186cb037a55324020e25dc222adea634d34d7c0502e1741502e585f1d104 · at the commit
test_sensitivity.py Sensitivity unit and integration tests · 7.7 KB · SHA-256 85a18170b3d1542372d657a8f58e1949455cfbce16f31d9c1ba54cdb0532c6bf · at the commit

Manifests

artifact_hashes.json SHA-256 inventory of the sensitivity result files · 823 B · SHA-256 3109998c6dd2d1d40fee6801b1370062d869f6e9af9e9e36d873ee61b958ce8f · at the commit

Protocols

protocol.json Sensitivity protocol (frozen before the new calculations) · 4.6 KB · SHA-256 7f22f7a108501bd0cccbd24e2a0fe76d6feec75ba118cdc1bbf9d17824fc1b78 · at the commit
protocol_amendment_01.json Protocol amendment 01 (recorded before execution) · 1.3 KB · SHA-256 88e965e5e0e98517e7dbb8149a1b94dd171d25da84070f8f45e06cc78fbb9d01 · at the commit

Logs

test_sensitivity.log Sensitivity test run log (11 tests) · 1.4 KB · SHA-256 b74bb793d67d23d6b45dc8b2112ee2d45f631a32590b4781412eb52ffbfc7ce7 · at the commit

Statuses

Archived result: A value read from the frozen October 17 evidence snapshot or from a result file versioned in the research repository. Nothing was recomputed here.
Executed reanalysis: A real run of the original pipeline with an execution record: date, interpreter and libraries, every input verified by SHA-256, and each output compared with the archived SHA-256.
Exploration: An interactive view over public data. It computes no scientific result and is not evidence.