A Hierarchical Probabilistic Formulation and a Mesoscopic Analytical Level for Cancer Therapy

Modern cancer therapy is investigated primarily through micro-level analyses of molecular, cellular, metabolic, immune, microenvironmental, and physiological mechanisms, together with macro-level evaluation of observable therapeutic outcomes. Current approaches commonly bridge these levels by relating selected microscopic features to outcomes, but they do not explicitly represent the collective dynamic process through which therapeutic outcomes emerge. Recent work formulated cancer therapy as a continuous process in which an initial state-space organization θ_0 and therapeutic input μ give rise to therapeutic trajectory evolution T, attractor formation A, and an observable outcome y: (θ_0,μ) → T → A → y. The present work examines the structural limitation of direct micro-to-macro mapping and current feature-based approaches, and then uses the probability chain rule to factorize the joint conditional distribution P(T_i,A_j,y_k ∣ θ_0,μ), where T_i, A_j, and y_k denote possible realizations of therapeutic trajectories, attractors, and outcomes, respectively. This factorization makes explicit the hierarchical conditional organization of therapeutic evolution. The continuous-process formulation and its probabilistic decomposition identify an analytical domain in which state-space organization, trajectory evolution, and attractor formation are treated as collective dynamic objects; this domain is termed the mesoscopic analytical level. By explicitly representing the therapeutic process between treatment and outcome, the framework formulates a dynamic probabilistic therapeutic-process mapping distinct from a direct input–outcome mapping. Coarse-graining is introduced as one possible methodological principle for transforming high-dimensional microscopic information into lower-dimensional representations suitable for mesoscopic analysis.