Stability‐guided transition modelling reveals an Active‐to‐Evasive immune ecosystem continuum in cervical cancer

Abstract Background The transcriptomic organization of tumour immune ecosystems in cervical cancer remains poorly defined beyond conventional static immunophenotyping frameworks. Methods In the present study, we developed a stability‐guided and transition‐aware systems immuno‐oncology framework to characterise immune ecosystem organisation in TCGA‐CESC. Results This framework identified a biologically coherent Active‐to‐Evasive immune ecosystem continuum through continuous transcriptomic modelling, stability analysis, and complementary immune ecosystem characterisation. Active tumours demonstrated preserved interferon‐associated and cytotoxic immune programs, whereas Evasive tumours exhibited coordinated immune attenuation accompanied by hypoxia‐associated and adaptive ecosystem‐remodelling programs. The Intermediate subgroup occupied a biologically structured region between the Active and Evasive ends of the continuum, characterised by graded attenuation of immune‐active programs, altered immune‐gene coordination, and transition‐associated transcriptomic remodelling. Bootstrap‐guided repeated‐convergence analyses identified reproducible transition‐associated sentinel genes, including SIRPG , TIGIT , CD96 , PRF1 , NKG7 , and TRAC , which consistently converged across trajectory‐associated analyses, stability profiling, Evidence Matrix prioritisation, biologically anchored compression, and complementary analytical layers. Single‐cell localisation analyses demonstrated preferential enrichment of these sentinel genes within activated cytotoxic immune compartments, whereas LM22‐based deconvolution supported graded immune compositional remodelling across the continuum, including reduced CD8 T‐cell and M1 macrophage‐associated components together with relative enrichment of neutrophil‐ and macrophage M0‐associated fractions towards the Evasive end. Sensitivity analyses further demonstrated that the underlying continuum was retained after outcome‐independent reconstruction, exclusion of immune‐definition genes, and adjustment for tumour purity. Conclusions These findings support a cross‐sectional systems immuno‐oncology model in which cervical cancer tumours are organised along an Active‐to‐Evasive immune ecosystem continuum characterised by graded immune attenuation, adaptive transcriptomic remodelling, and increasing immune–tumour uncoupling. The Intermediate ecosystem state represents a biologically structured transition‐associated region rather than a simple residual category, whereas the compressed transcriptomic representation captured a parsimonious component of the broader continuum architecture and was interpreted as a biologically anchored projection rather than an independently validated prognostic signature. Highlights A stability‐guided framework characterises an Active‐to‐Evasive immune ecosystem continuum in cervical cancer. The Intermediate state represents a biologically structured, transition‐associated region of the inferred continuum. Sentinel genes and complementary cellular analyses support coordinated immune remodelling across ecosystem states. A biologically anchored 7‐gene compression preserves core features of the broader continuum architecture.