Cell composition accounts for two-thirds of the salivary gland transcriptome signal in primary Sjögren's syndrome: analysis code and derived tables

Analysis code and derived tables for a pre-specified triangulation of bulk expression, single-cell pseudobulk and eQTL Mendelian randomisation in primary Sjögren's syndrome.The archive contains the analysis pipeline, one script per layer (sample cleaning and gene-space harmonisation; differential expression and gland-versus-blood classification; composition scoring and calibration against pseudo-bulk ground truth; single-cell extraction, pseudobulk differential expression and module scores; eQTL instruments, Mendelian randomisation and MHC masking; cross-tissue concordance with permutation nulls), together with every derived table cited in the manuscript (97 files).The following numbers are reproducible directly from the deposited tables: 2,514 genes at FDR<0.05 in salivary gland and 759 in blood within the 15,638-gene space (579 gland-specific, 168 shared); calibrated composition scores explain 62% of the gland signal (range 45-96% across four scoring schemes); the only cell-intrinsic programme surviving correction is CD4 T-cell activation/exhaustion (d=2.60, FDR=1.0e-4), and interferon-stimulated genes show no elevation in any of 15 cell types; after MHC masking, eQTL-based Mendelian randomisation implicates 2 salivary-gland and 8 blood genes, none of them differentially expressed.Raw data are not redistributed. They are public: GEO GSE40611, GSE23117, GSE66795 and GSE18732; GTEx v8 via the eQTL Catalogue; FinnGen release 13; single-cell data via CZ CELLxGENE Discover (collection 21bbfaec-6958-46bc-b1cd-1535752f6304). The scripts point to a local mirror; adjust the root path constant at the top of each script.The corresponding manuscript is under review; this record will be linked to it once published.