Determinability of cell-type-specific effects in major depression: lightweight analysis archive

This archive contains analysis and plotting code, full gene-level differential-expression results, summary tables, a result registry, a data dictionary, source and exclusion manifests, and provenance metadata for an audit of cell-type-specific claims in public post-mortem single-nucleus data on major depression. It is a research-data archive; journal manuscripts, review records, and rendered figures are not included. The retained tables cover 51 donor-level RNA comparisons based on 71 coded RNA donors and an ATAC analysis based on 83 coded donors. Across the reported female and male comparisons, the cell-level analyses yielded 40,163 significant gene–comparison calls, compared with 104 donor-level calls. This difference is a contrast between counts, not an empirical false-discovery-rate estimate. For 29 of 51 comparisons (56.86%), the minimum detectable effect exceeded the reported source-study median absolute effect of 1.224 log2-fold change. Six empirical permutation p-values were below 0.05 before correction across comparisons. The adjusted ATAC analysis reported no significant peaks across 4,293,716 peak tests. The package includes a standard-library verification script for the principal arithmetic claims. Third-party raw data, donor-level metadata, and two large rebuildable pseudobulk matrices are not redistributed; their sources, sizes, SHA-256 hashes, and access or rebuilding routes are recorded in the manifests. A complete rerun from original GEO files has not been independently validated. Donor non-overlap between source studies has not been verified. Derived data and documentation are licensed CC BY 4.0; code is provided under the MIT License as detailed in the package LICENSE file. Correspondence: Cheng Zhu, [e-mail].