The mouse 39 °C iTreg anchor — GSE329522

Bulk RNA-seq of induced regulatory T cells cultured at 37 °C and at 39 °C, in wild-type and cGAS-knockout genotypes.

This is the mouse half of a project asking whether human inflammatory and autoinflammatory disease states carry transcriptional programs consistent with a temperature-stress axis in T cells and with cGAS–STING-related biology. It derives the signatures; the human compartments score them.

Design

Induced regulatory T cells differentiated from primary murine splenic CD4⁺ T cells, in a 2×2 design of genotype (wild-type, cGAS-knockout) × temperature (37 °C, 39 °C), five biological replicates per group, twenty libraries. Deposited as GSE329522.

Seven contrasts follow from that design. The analysis fits them, reads the rankings against thirteen gene-set databases and two regulatory networks, freezes the genes that rise at 39 °C into named signatures, converts those to human symbols, and decomposes them against curated lenses.

Findings

• Temperature dominates. PC1 carries 57.3 % of the variance and tracks temperature. Warming changes 8,723 genes in wild-type and 8,901 in cGAS-knockout cells at FDR 0.05.
• Both genotypes warm in step. Across 10,418 heat-responsive genes the knockout response is 0.99× the wild-type response at r = 0.95.
• The cGAS-dependence test is the weakest contrast — the heat-by-genotype interaction, at one degree of freedom and n = 5, where 23 genes clear FDR 0.05. Every artifact here phrases a failure there as no detectable cGAS-dependence at n = 5.
• Interferon and hypoxia diverge under that test. The two Hallmark interferon-response sets carry the interaction at +3.14 and +2.83, adjusted p 2.7e-24 and 5.0e-22. HALLMARK_HYPOXIA rises in both heat arms at +1.91 and +1.95 and stays flat in the interaction at −1.27, adjusted p 0.17.

Each number is reproduced in a table under 03_results/, named in that stage's README.md.

The deliverable

03_results/human_projection/ is the contract the human compartments read — three frozen arms in human symbols, each with an up list, a down list, a signed ranked list, and a manifest of sizes and gates. Genes are gated on FDR and |log2FC| ≥ 1 and converted with pinned offline babelgene 22.9, with a ledger recording every gene's fate through that step.

Signature	Mouse symbols	Human symbols
WT_heat_up	213	202
KO_heat_up	239	221
Interaction_up	9	7

Conventions

Signs. A positive statistic points to the numerator of its contrast: 39 °C for a heat contrast, wild-type for a genotype contrast. For the interaction, positive means the heat response is larger in wild-type.

Naming. A signature is named for how it was derived, so WT_heat_up is checkable against the differential-expression tables. A gene set enriching locates its gene content in a ranking; whether the program it is named for is present is measured separately, in the decomposition stages.

Layout

Path	Contents
00_data/	The normalized CPM matrix with its provenance record, and the symbol-alias map. Scripts fetch or cache the external reference sets; sequence data lives in GEO.
02_analysis/scripts/	60 numbered scripts, 59 in R and one in Python. Compute scripts write tables; _viz scripts draw from them.
02_analysis/config/	analysis_config.yaml — every path and threshold the scripts read.
02_analysis/helpers/	Figure styling, ortholog conversion, symbol-alias resolution, source hashing.
03_results/	One directory per stage, each with tables/, figures/ and a README.md captioning every file.

03_results/README.md is the entry point: reading order stage by stage, each finding with the table it comes from, and the provenance of every gene set, lens and network the analysis reads.

Reproducing

Scripts run in numeric order inside the container under .devcontainer/, reading paths and thresholds from 02_analysis/config/analysis_config.yaml. Tables and stage READMEs are the tracked record — figures regenerate from the script each caption names, since .png and .pdf are ignored repo-wide. Clone with --recursive to populate 01_modules/, which GitHub source tarballs carry as empty directories.

Environment

The analysis runs in scdock-r-dev:v0.5.10, pinned on the dev-core service in .devcontainer/docker-compose.yml. That image is defined by scbio-docker at commit 5885cd3.

RNAseq-toolkit supplies the GSEA plotters and the prebuilt mouse gene-set objects — TransportDB, MitoCarta 3.0 and MitoXplorer 3.0 — read from its reference tree. This release records it at commit 752481f (v0.2.0-9-g752481f, on dev) under 01_modules/.

License

MIT for code — 02_analysis/, .devcontainer/, and scripts anywhere in the tree (LICENSE). CC BY 4.0 for results and prose — the tables, figures and README text under 03_results/, and the documentation (LICENSE-CC-BY-4.0.txt). Both require attribution. Sequence data carries the terms of its GEO deposition.