Pan-cancer analysis of context-dependent associations between SASP-related transcription and inflammatory programs
Introduction Bulk transcriptional scores related to the senescence-associated secretory phenotype (SASP) are indirect measures and do not directly quantify SASP protein secretion or senescent-cell abundance. Methods We analyzed SASP-related transcription in 9,698 primary tumors from 32 TCGA cancer types and examined external associations in two GEO datasets: GSE39582, represented by discovery and validation subsets from the same study, and GSE30219. TCGA used a fixed Reactome-derived 102-gene ssGSEA score. GEO used a 27-gene core score and a 20-gene sensitivity score, both calculated as means of gene-wise z scores. Cancer-specific associations were adjusted for ABSOLUTE tumor purity and summarized using random-effects meta-analysis, with direct-overlap controls and composition-sensitivity analyses. Results TCGA pooled inflammatory correlations were positive but modest (ρ = 0.128–0.258), with substantial between-cancer heterogeneity (I 2 = 87.9–91.5%); stromal-context associations were weaker and more variable. GEO pooled inflammatory correlations were larger (ρ = 0.641–0.819), but differences in exposure signatures, scoring, platform and analytical settings prevented attribution of this discrepancy to a single factor. All four GEO DerSimonian-Laird prediction intervals crossed zero. REML/Hartung-Knapp-Sidik-Jonkman confidence intervals for the pooled means excluded zero, but only two source datasets were represented. Exploratory matched random gene-set benchmarks did not show that SASP102 correlations exceeded the specified random expectations. In GSE39582, core scores were higher in proximal and dMMR tumors and in C2/C4 molecular-subtype contexts. Post-review OS pooling using stage-stratified GSE39582 models remained non-significant (27-gene score: HR = 1.001, 95% CI 0.896–1.119); conventional validation-subset fits had convergence warnings. Discussion SASP specificity and consistent prognostic utility were not established; the findings support context-dependent inflammatory transcriptional associations under the predefined scores.
Authors
- Lei Gao (ORCID: https://orcid.org/0000-0003-1476-1460)
- Rong Liu (ORCID: https://orcid.org/0000-0001-5170-6474)
- Wenliang Chen
- Zhao Li
- Fei Zhang
- Haozhi Qin
- Fei Han
Institutions
- Shanxi Medical University (CN)
- Second Hospital of Shanxi Medical University (CN)
Publication Details
- Journal
- Frontiers in Bioinformatics
- Published
- 2026-11-16
- DOI
- https://doi.org/10.3389/fbinf.2026.1975401
- Primary Topic
- Telomeres, Telomerase, and Senescence
- Type
- article
- Field-Weighted Citation Impact
- 0.00