From binary senescence to state-resolved senotherapy in cancer: therapeutic windows from heterogeneity.

Tumor senescence is a durable cell-cycle arrest triggered by oncogenic signalling, DNA damage and therapeutic stress. Although senescence can restrain malignant expansion, heterogeneous senescence-associated secretory programs can also promote tumor progression, immune evasion and treatment resistance. Crucially, the composition, magnitude and persistence of these secretory programs vary across cell types, microenvironmental niches and treatment phases, making binary detection of "senescent cells" insufficient for deciding whether specific populations should be eliminated, modulated or preserved. This review synthesizes the molecular determinants of functional, temporal and spatial heterogeneity in tumor senescence and consolidates them into an operational state space to support phase-aware intervention logic. It further evaluates how artificial intelligence, combined with single-cell and spatial profiling, imaging and circulating readouts, can enable state-resolved mapping, stratification and monitoring of senescence contexts, thereby generating testable therapeutic window hypotheses and guiding the development of staged precision senotherapies.