Analysis package for A detectability floor for treatment-response labels in rheumatology transcriptomics
Year: 2026
Type: software
Abstract: Deposit for the manuscript "A detectability floor for treatment-response labels inrheumatology transcriptomics". The archive contains everything needed to reproduce thereported numbers from public input data; no new data were generated and no controlled-accessmaterial was used. Contents: data provenance for the extended label audit (platform, annotation priority, genecoverage, retrieval date, label field and mapping rule, deposited-versus-published group-sizecomparison); the pre-registered audit protocol; per-sample response labels recovered from eachdeposit; the screening and adjudication ledgers taking 45 auditable papers down to 15 auditedcohorts, with a registered reason for every exclusion; the analysis and plotting code for theprimary cohorts (juvenile dermatomyositis, rheumatoid arthritis before/after methotrexate,lupus nephritis, two controlled glucocorticoid challenges, a single-cell dose series), thegender-composition and per-gene stability checks, the score-level injection calibration andthe attainable-AUC ceiling analysis; and the corresponding result files. Design floors are exact non-central-t minima (80% power, two-sided alpha = 0.05) computed foreach cohort from its own group sizes, so a negative result is reported as a bound rather thanan absence. Verdicts follow a taxonomy that was registered before any verdict was computed.A verdict characterises the detectability of a supervision label as deposited; it is not aclaim that any published analysis was erroneous. No virtual-cell model was evaluated in thiswork. Language convention: file names, field names, column headers, verdict codes and all numericoutput are in English. Recorded screening reasons, adjudication notes and analysis logannotations are in the authors' working language (Chinese); the full PRISMA-style screeningrecord is deposited in its original form with an English summary prepended, and every ledgeris described in ledgers/README_ledgers.md.
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7219607396