Mitoxyperilysis-ARDS: analysis code, six-layer gene signature, and frozen statistical outputs for "Mitoxyperilysis-consistent L4 de-protection in human sepsis-associated ARDS blood: a compartment- and etiology-bounded transcriptomic boundary map" (archive v6; v2.34)
This archive supports the manuscript "Mitoxyperilysis-consistent L4 de-protection in human sepsis-associated ARDS blood: a compartment- and etiology-bounded transcriptomic boundary map" (submission target: Critical Care, BMC; manuscript version v2.34, September 2026). Contents: (1) complete analysis code (R 4.6.1 / Python 3.11) covering score construction (ssGSEA/UCell), four-cohort meta-analysis with Hartung-Knapp inference, leave-one-gene-out robustness, three-method deconvolution, single-cell localization (Scissor, diffusion pseudotime), transcription-factor activity inference (decoupleR/DoRothEA, PROGENy), Mendelian randomization (eQTLGen instruments; ARDS GWAS GCST90700332; FinnGen R13 endpoints), pharmacological layers (DSigDB, CMap/LINCS, AutoDock Vina docking, DGIdb/Open Targets), and permissive-state cross-disease comparisons; (2) the six-layer mitoxyperilysis gene signature (GMT/CSV); (3) frozen statistical outputs (module CSV files plus eTables) serving as the audit anchor for every statistic reported in the manuscript; (4) data-lineage documentation, the pre-specified analysis plan, and contemporaneous analysis logs. Changes in this version relative to version 5: version 5 was followed by an R52 wording correction (the manuscript's Data Availability had cited this archive as "Zenodo; 2026 (version 3)" while giving the concept DOI 10.5281/zenodo.22135916 as the accession; the fixed label could contradict the concept DOI, which always resolves to the latest version, so the label was removed) and by an R53 independent quality-control repair round (manuscript v2.31 to v2.32, wording- and documentation-level only): one arithmetic token in Results (the benchmark chance expectation, ≈0.9 → ≈0.875 = 7 × 0.125); one methodological sentence on the Hartung–Knapp variance-inflation factor added to the Statistical analysis section; the GSE326212 donor-threshold tokens in eMethods item 13(g) corrected (≥20 → ≥30 neutrophils per donor) with the provenance and SHA-256 of the GSE326212 donor-level h5ad recorded in the same item. In the results archive, two eTable CSVs received wording-level repairs (eTable1: the GSE326212 source caliber now distinguishes the 17 GEO sample entries from the authors' aggregated 64-donor h5ad cohort; eTable8: one description row translated into English) and the eTables README gained three annotation notes (a 0-byte placeholder output, a superseded weighted-Stouffer value, and a pre-redefinition L5 caliber note); no numeric content was changed. An R54 final pre-submission quality-control round (2026-09-19, v2.32 to v2.33) compressed the abstract to the journal limit, restructured one Results paragraph, and added a Table 2 multiplicity note and provenance/boundary sentences; no statistic, direction, frozen output value, or conclusion changed. An R57 round (2026-09-20, v2.33 to v2.34) reworded two passages — the sepsis-prognosis boundary sentence, now reporting the measured outcome-annotation gap, and a Methods note on the non-interchangeability of the FinnGen _WIDE endpoints — again without altering any statistic, direction, frozen output value, or conclusion. An R61 round (2026-09-20) was engineering-only: it made the bundled verifier and archive builder independent of the manuscript version and rebuilt this archive from the current sources, adding no analysable output and changing no frozen value. code.zip carries the script families added since version 5 (R52, R53, R57, R61), together with the version-agnostic verifier and the current builder; metadata.zip carries the data-lineage document with the registration sections added since version 5 (R52 through R61). analysis_plan.zip and supplement_matrices.zip are byte-identical to version 5; frozen_outputs.zip differs only in the three wording-level files under eTables/ named above, plus one donor-level table newly supplied under module8_crossdisease/ (the status-corrected 52-donor intermediate behind eTable 5a, recomputed from previously frozen cell-level scores); every pre-existing file is byte-identical to version 5. No reported statistic, effect direction, frozen output value, or conclusion differs between version 5 and this version. Relative to version 3, this archive remains the extended set: the frozen-output set and analysis code cover the R35 extensions (sepsis-prognosis survival analysis in GSE65682; protein-layer coverage screen including pQTL colocalisation and Mendelian randomisation for the HMGCR-RhoA axis; structured plasma/phospho/ATAC screens), the R37 analyses, the R39 Task F.1 NETosis orthogonality benchmark, the R40 sensitivity-endpoint analyses, and the R49/R50 documentation and verification updates. The data-lineage document in metadata.zip has been updated accordingly (raw-archive manifest now 228 entries). A read-only metadata verifier for the associated OSF registration is included in the code archive. All frozen values reported in manuscript versions up to v2.34 are represented. Title correspondence: earlier manuscript versions and the associated OSF registration (osf.io/p3wcy, registered 2026-08-28) carry the former title "...in human sepsis-associated ARDS neutrophils: a shared stress program constrained by compartment and etiology". The first revision was wording-level (the compartment claim was narrowed to blood; the subtitle was removed) and altered no data, statistic, direction, or conclusion. The title was subsequently revised twice more, again without any change to the data, statistics, effect directions, frozen outputs, or conclusions: at v2.28 the headline claim was re-scoped from a uniform blood signature to a compartment- and etiology-bounded boundary map, with the composite MAS score moved to a secondary readout in the abstract and the L4 arm named as the discriminating signal; and at v2.29 the modality name "mitoxyperilysis" was restored to the headline, with the v2.28 boundary-map framing retained as the subtitle. The OSF registration at osf.io/p3wcy was not retro-edited; the divergence between the submission title and the registered title is disclosed in the registered amendment osf.io/2qkg3 (doi:10.17605/OSF.IO/2QKG3, 2026-09-12) and in the manuscript itself. All primary datasets are publicly available (GEO accessions GSE66890, GSE32707, GSE10474, GSE76293, GSE235046, GSE200848, GSE326212, GSE145926, GSE151263, GSE167363, GSE188288, GSE65682; PRIDE PXD060437; Metabolomics Workbench ST002738; GWAS summary statistics via GWAS Catalog GCST90700332, FinnGen R13 and eQTLGen official portals) and are NOT redistributed here. GWAS summary statistics are not redistributed because FinnGen and eQTLGen distribute them under their own access terms; use the official portals. Version history and correction: this record is the sixth version of a continuously versioned archive. The record published as version 4 (10.5281/zenodo.22736755, 2026-09-13) was created from a new-version draft that inherited the version-3 file set; that inherited file set was not removed before upload, so version 4 did not contain the extended archive described above and has been superseded. Version 5 (10.5281/zenodo.22738936, 2026-09-13) carried exactly the file set intended for version 4: the complete analysis code and the full frozen-output set, together with the updated data-lineage document. This version (v6) changes the code, metadata and frozen-outputs archives (the last only in three wording-level files under eTables/), as described above, and leaves analysis_plan.zip and supplement_matrices.zip byte-identical to version 5. The concept DOI 10.5281/zenodo.22135916 always resolves to the latest version.