Drought-associated transcription of raffinose-family oligosaccharide genes in tomato: an exploratory study-level RNA-seq meta-analysis

Reproducible meta-analysis of drought response in tomato (Solanum lycopersicum), focused on the raffinose family oligosaccharide (RFO) / galactinol pathway. 63 candidate BioProjects were screened; 27 were reanalysed from raw reads (30 strata, 435 libraries), and two further BioProjects (PRJNA1380757, PRJNA931112) were held out as an out-of-sample replication cohort. Quantification used salmon 2.8.0 (k = 31) against the decoy-free ITAG4.1 cDNA reference. The primary cohort is 13 concurrent-control leaf/shoot strata, 66 control and 73 drought libraries, 21,967 genes tested. Every table and figure here comes from one sealed end-to-end pipeline run: 2026-11-01, 20:01:59–21:56:36 UTC, git commit b6319cb, 44 tracked code and configuration files, 61 inputs, 174 outputs, exit code 0. The seal step refuses to complete if any tracked file changes while the run is in flight. tomato-drought-frozen-tables.zip contains the run manifest with per-stage commands, timings, exit codes and SHA-256 hashes of every input and output. Contents tomato-drought-count-matrices.tar.gz — gene-level count matrices and sample sheets for all 30 strata (61 files) tomato-drought-meta-pipeline.tar.gz — the complete analysis pipeline as run tomato-drought-frozen-tables.zip — 43 frozen result tables and the run manifest tomato-drought-figures.zip — the 8 figures tomato-drought-run-logs.zip — 20 stage logs tomato-drought-curation-and-provenance.zip — curation notes for every BioProject, genotype-arm notes, the replication pre-registration, data-feasibility report, quantification QC logs and curated sample tables 00_ICINDEKILER.md, 00_dosya_envanteri.csv — folder guide and SHA-256 inventory of all 118 files Main result The RFO biosynthetic arm (7 testable genes) has a mean pooled log2 fold change of +1.34 across the 13 concurrent-control strata, against an expression-matched background null of −0.32 to +0.37; the variance-inflation-corrected empirical p is 0.0028, and the correlation-aware rotation test combined across strata gives p = 0.00047. A related glycosyltransferase control set (GT8) sits inside its own null (−0.05, p = 0.94). Limits Of the 19 testable panel genes only one galactinol synthase (Solyc02g084980) clears genome-wide FDR, and under the HKSJ variance estimator no gene does. Panel, cohort and exclusion decisions were all fixed after the results were seen, so no p-value computed inside this corpus is a calibrated error rate. Sample labels, control harvest times and biological-replicate independence rest on archive metadata and have not been verified against the originating laboratories. The calibration simulation tests specific synthetic conditions and does not calibrate very small p-values. Not included Manuscript text and writing/submission documents are deliberately not deposited here, nor are the external audit reports or the prose-check logs that quote manuscript sentences. Nothing excluded produces a result; the deposit regenerates end to end on its own.