Context-Aware Modelling of the Senescence-Associated Secretory Phenotype: A Public Proteomics Benchmark Across Inducers, Cell Types, Compartments, and Time

The senescence-associated secretory phenotype (SASP) is widely used to identify senescent cells and to motivate senomorphic and senolytic interventions, yet its composition varies with cell type, inducing stress, time, and secretory compartment. We present a reproducible secondary benchmark for context-aware SASP modelling using the CC0 SASP Atlas proteomics resource. Eight soluble or extracellular-vesicle contexts were reanalysed using the original significance rule (q<0.05 and fold change>1.5), yielding 2,446 quantified and 1,295 elevated protein groups in the union. The number of elevated factors ranged from 21 to 709 per context. Pairwise Jaccard overlap was low (median 0.062), 785 of 1,295 elevated groups (60.6%) occurred in only one context, no group was elevated in all eight contexts, and only four occurred in six. Overlap was higher among late soluble fibroblast SASPs generated by different inducers (mean Jaccard 0.257) than between matched soluble and extracellular-vesicle compartments (0.047), matched early and late time points (0.146), or irradiated fibroblast and epithelial SASPs (0.062). In leave-one-context-out protein prioritisation, simple prevalence, mean-effect, stability-adjusted, and metadata-kernel scores achieved median average precision of 0.173, 0.220, 0.215, and 0.221, respectively. The modest gains over context prevalence and the low recall of small panels show that apparent universal markers transfer incompletely to new biological settings. Finally, a non-compensatory analysis combining worst-case SASP induction across irradiation, oncogenic RAS, and atazanavir with replicated plasma age association identified GDF15, MMP1, and TIMP1 on the Pareto frontier. These results support a two-level strategy: context-conditioned SASP models for detection and response modelling, plus separately validated cross-context panels for translational biomarker use.