Use the Allen Immunology aging cohort PBMC scRNA-seq data (multi-donor, age-stratified) with linked CMV serostatus metadata. Run pseudobulk DESeq2/limma-voom models on CD8 subsets (Tex-prog, Tex-int, Tex-term, TEMRA, GZMK+ effector-memory) with models: (1) age + sex; (2) age + sex + CMV; (3) CMV alone. Report variance partitioning (variancePartition R package) to estimate CMV-attributable vs. age-attributable variance per gene and per subset. Secondary: compositional analysis (miloR neighborhood differential abundance) with and without CMV covariate. Success criterion: ≥1 CD8 subset where CMV adjustment shifts age effect size by >25% or flips direction. This will directly inform which ‘aging’ CD8 signatures are CMV-dependent vs. CMV-independent and should be a mandatory covariate check for any Allen cohort aging claim.
Details
- disease
- immune aging, immunosenescence, inflammaging
- target_ref
- Allen Immunology aging cohort PBMC scRNA-seq; CMV serostatus metadata
- primary_endpoint
- Gate issues a binary PASS or FAIL (not BLOCKED): requires ≥2 papers with numeric evidence for both TOX+PDCD1+ exhaustion and CD57+KLRG1+ senescence subsets in aged donors (≥60y)
- identification_strategy
- observational
Raw fields (3)
- assay_spec
Literature triangulation: extract quantitative CMV×age×CD8-exhaustion separability metrics from Kared 2024 (PMID 39730488), Mogilenko 2021, and Terekhova 2023 via PubMed/EuropePMC full-text; tabulate SLAMF7, GZMK, GZMB, TOX, PDCD1, CD57, KLRG1 marker co-expression by age stratum and CMV status; issue pass/fail gate for exhaustion-vs-senescence separability; Tabulated marker co-expression evidence: for each paper, report the quantitative metric (proportion, logFC, p-value, or threshold) for TOX+PDCD1+ (exhaustion) vs CD57+KLRG1+ (senescence) CD8 subsets, stratified by age decade and CMV status where available — planned; {'kind': 'claim', 'text_template': 'Exhaustion-vs-senescence separability [PASSES / FAILS / BLOCKED] the numeric gate: [≥2 papers report distinct TOX+PDCD1+ and CD57+KLRG1+ populations with non-overlapping age trajectories after CMV adjustment] — planned'}; Explicit blocker statement if gate cannot be issued: naming which of CMV status metadata, clonotype expansion data, or per-donor model output is absent from all three sources — planned- hypothesis
A substantial fraction of the age-associated CD8 T cell transcriptional and compositional variance attributed to 'aging' in PBMC cohorts (including GZMK+ inflammaging and TEMRA expansion signatures) is confounded by CMV serostatus, and this confounding is not uniformly adjusted for across published aging cohorts. Stratifying Allen Immunology aging cohort donors by CMV IgG serostatus and including CMV status as a covariate in pseudobulk limma-voom models will materially alter the effect-size estimates for age-decade on CD8 subset composition and exhaustion-marker expression (PDCD1, HAVCR2, TOX, GZMB, KLRG1, CD57/B3GAT1).
- kill_criteria
Abort or revise if: Mogilenko 2021 may not report CMV-adjusted models — gate may be BLOCKED on CMV covariate rather than marker separability; Terekhova 2023 DOI unverified; may require OpenAlex fallback search to retrieve full text; Kared 2024 reports SLAMF7 subsets but may not cross-tabulate with TOX/PDCD1 exhaustion markers directly — separability inference would then be indirect