One promising advantage of senolytic agents is to selectively induce programmed death of senescent cells, such as ABT-263, ABT-737 and the combined use of dasatinib and quercetin11,18,19. We first tested the efficacy of these geroprotective drugs against senescent PSC27 cells to demonstrate its potential as an experimental cell model for drug screening. Our preliminary data suggested that each of these compounds significantly depleted senescent cells but not proliferating cells, thus confirming the feasibility of using this stromal line for further studies (Extended Data Fig. 1b). Upon large-scale screening of the PDMA library, we identified several compounds with the potential to selectively kill senescent cells in culture (Extended Data Fig. 1c–e). Among the agents showing preliminary anti-senescence effects were GSE, quercetin, fisetin, curcumin and piperlongumine (Extended Data Fig. 1d,e). Quercetin and fisetin share similar chemical structures, exert similar medicinal effects and are both known senolytics11,20,21. Curcumin and piperlongumine are also natural compounds with recently discovered senolytic potential22,23. We chose to focus on GSE, which remained a largely underexplored source. Under in vitro conditions, GSE suppressed the SASP with maximal efficiency at 0.1875μgml−1 (Extended Data Fig. 2a), which fits with the property of senomorphics24. Lower or higher concentrations of GSE were less efficacious, perhaps due to the induction of cellular stress responses as a result of increased cytotoxicity (ExtendedData Fig. 2a). Using RNA-seq, we found that treatment with GSE significantly altered the expression profile of senescence cells, with 2,644 genes downregulated and 1,472 genes upregulated at a fold change of 2.0 per gene (P