Provenance constrains mitochondrial transcriptomic inference in autism cortex

Reproducibility package for the provenance audit and mitochondrial case study reported in “Provenance constrains mitochondrial transcriptomic inference in autism cortex.” Version 3.5.0 reconstructs the Liu 2016 publication cohort from GSE51264, GSE59288, and supporting files, then restricts the primary reconstructed-matrix contrast to 25 ASD and 28 control donors within observed lower-age support. All ASD profiles originated in GSE59288 and all controls in GSE51264, leaving diagnosis inseparable from source accession. The coefficient therefore cannot identify an ASD disease effect separately from accession-specific processing. Official metadata document different deposited processing pipelines, and the source-to-publication correlation is 1.000 for ASD profiles versus a median of 0.744 for controls. Source assignment is invariant across all-gene, variance-stratified, rank-based, and random-subset analyses. These observations establish asymmetric processing without identifying a separate accession coefficient. The package contains ten MitoCarta-derived modules, complete gene membership and hierarchy strings, HC3 standard errors, descriptive 90% confidence intervals, leave-one-donor-out diagnostics, RIN nonlinear and overlapping-range analyses, near-tied source-metadata perturbations, pseudocount and scoring sensitivities, subgroup distribution tests, respiratory-complex and sub-mitochondrial endpoints, a complete 84-to-79-to-67 Anitha/QIAGEN panel mapping audit, analysis-history disclosure, donor-independence checks, competitive gene-set tests, and independent software numerical validation. GSE102741 is retained as an independent low-precision comparison. Its external ASD cortical composition/quality reference-axis directions were concordant but imprecise; these axes are not a positive control for an identifiable ASD effect. The cohort is reported separately and is not pooled with Liu. Third-party source datasets are not redistributed. The archive contains source URLs and SHA-256 checksums, frozen mappings, numerical results, three main and eight supplementary figures, tables, environment files, manuscript snapshots, a 58-reference validated bibliography, the Brain Research graphical abstract, the frozen Gandal 2022 reference-axis TSV with provenance metadata, network-independent V1–V8 statistical tests, CITATION.cff, and integrity manifests. Frozen results, Supplementary Table S2, and the file manifest permit offline checking of all reported numbers without downloading the third-party source datasets. Code is MIT licensed; author-generated documentation, tables, and figures are licensed under CC BY 4.0.