{
"dataset": {
"ref": "dataset:b5c92e78-3596-432b-ab17-f3306c7d0dd4",
"accession": "GEO GSE271896",
"data_type": "pseudobulk_scrna_seq",
"size_note": "Use pseudobulk h5ad and clinical CSV — NOT the 87 GB full h5ad — to stay within per_cycle budget cap.",
"description": "Allen/Sound Life Project scRNA-seq dataset (Gustavson et al. Nature 2025). 1,094 donors, 2,638 samples. Per-sample Cell Ranger outputs at GEO GSE271896; pseudobulk h5ad and clinical metadata CSV at Allen AIFI portal (doi:10.57785/e9e1-wh09). Access: public, no controlled-access gate."
},
"objective": "Test whether CD8+ T cell exhaustion score (TOX/PD-1 co-expression at pseudobulk level) is a statistically significant negative predictor of post-vaccination antibody titre fold-change in older adults, after adjusting for known confounders. This directly evaluates hypothesis hypothesis:2bf795a4-6b9e-4bc7-a1c2-9828347b63e8 using dataset dataset:b5c92e78-3596-432b-ab17-f3306c7d0dd4.",
"refs_json": {
"geo_gse271896": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE271896",
"pubmed_37845489": "https://pubmed.ncbi.nlm.nih.gov/37845489/",
"cellxgene_sound_life": "https://cellxgene.cziscience.com/collections/e9360edf-b0b7-4e01-bce8-e596814f13e7",
"gustavson_nature_2025": "https://doi.org/10.1038/s41586-025-09686-5",
"allen_immunology_portal": "https://apps.allenimmunology.org/aifi/insights/dynamics-imm-health-age/"
},
"authored_by": "persona-claire-gustavson",
"budget_note": "Uses pseudobulk h5ad only; full 87 GB h5ad not fetched — within per_cycle_author_tokens=5500 cap.",
"confounders": [
"CMV serostatus (major driver of CD8 exhaustion unrelated to aging per se)",
"Sex (immune-aging trajectories differ between females and males in AIFI data)",
"BMI (chronic inflammation mediator, confounds inflammaging signal)",
"Donor random effects (repeated samples per donor in longitudinal design)",
"Cross-sample batch effects (per Cell Ranger output batch from GSE271896)"
],
"cohort_strata": [
{
"role": "younger_reference",
"label": "young",
"age_range": "<50 years",
"rationale": "Baseline immune competence; expected high vaccine response"
},
{
"role": "transition_group",
"label": "middle",
"age_range": "50–64 years",
"rationale": "Tests for intermediate CD8 exhaustion signal before overt immunosenescence"
},
{
"role": "primary_study_group",
"label": "older",
"age_range": "≥65 years",
"rationale": "Primary stratum for vaccine efficacy assessment; predicted CD8 exhaustion high"
}
],
"source_policy": "public-safe",
"upstream_refs": {
"dataset": "dataset:b5c92e78-3596-432b-ab17-f3306c7d0dd4",
"landscape": "landscape:3202626b-e2f2-4823-8322-4707f947c729",
"hypothesis": "hypothesis:2bf795a4-6b9e-4bc7-a1c2-9828347b63e8",
"knowledge_gap": "knowledge_gap:c3a4edc9-06c7-46ba-ac66-d47f9b77c7fc",
"upstream_work_packet": "agent_work_packet:eb1bd02b-5585-48ea-8c1a-149f4aa7cb72"
},
"primary_metric": {
"name": "CD8_exhaustion_score",
"unit": "fraction_0_to_1",
"definition": "Pseudobulk fraction of CD8+ T cells co-expressing TOX and PD-1 per donor, normalised by total CD8+ T cell count. Measured at pseudobulk level from GEO GSE271896 Cell Ranger outputs.",
"measurement_level": "pseudobulk"
},
"cycle_iteration": 2,
"outcome_variable": {
"name": "post_vaccination_antibody_titre_fold_change",
"unit": "log2_fold_change",
"definition": "Log2 fold-change in antigen-specific IgG titre from baseline (Day 0) to Day 28 post-vaccination, from clinical metadata CSV."
},
"showcase_query_id": "immune-aging-vaccine-design-loop",
"statistical_model": "Linear mixed-effects model (LME): outcome ~ CD8_exhaustion_score + naive_B_cell_fraction + age_stratum + CMV_serostatus + sex + BMI + (1|donor_id). Donor random effect accounts for repeated measurements across the Sound Life longitudinal cohort. Propensity score matched sub-analysis in ≥65 stratum.",
"trajectory_run_id": "claire-arc2-20260525T035458-b6d1a76f",
"analysis_plan_steps": [
"1. Download pseudobulk h5ad and clinical metadata CSV from Allen AIFI portal",
"2. Extract TOX/PD-1 co-expression fractions per donor per cell type",
"3. Merge with clinical metadata (CMV, sex, BMI, vaccine response titres)",
"4. Fit LME model per age stratum; report β coefficients and 95% CIs",
"5. Propensity-score matched sub-analysis in ≥65 stratum",
"6. Sensitivity analysis: re-run excluding donors with CMV serostatus missing",
"7. Report competing predictor comparison: CD8 exhaustion vs. naive B cell loss"
],
"evaluation_criteria": [
"β(CD8_exhaustion) < −0.15, 95% CI excludes zero, adjusted p < 0.05",
"Competing hypothesis (naive B cell loss) tested in same model",
"Sensitivity analysis: result stable after CMV exclusion",
"Model diagnostics: residual normality, no influential outliers > 3 SD",
"Replication: qualitative consistency in at least one non-Sound-Life cohort (PMID:37845489)"
],
"falsifiable_prediction": {
"statement": "CD8 exhaustion score (TOX/PD-1+ fraction) negatively predicts post-vaccination antibody titre fold-change in the ≥65 stratum.",
"expected_direction": "negative",
"expected_effect_size": "β < −0.15 (standardised)",
"significance_threshold": "adjusted p < 0.05 (Bonferroni-corrected for 3 strata)",
"competing_hypothesis_test": "Include naive B cell fraction as co-predictor. If β(naive_B_cell_loss) dominates β(CD8_exhaustion) by >2-fold, primary hypothesis is not supported."
}
}