A Unified Ornstein–Uhlenbeck–Lévy–Branching Framework for Interpretable Modeling of Cancer Evolution
Biological evolution arises from the interplay of continuous stochastic dynamics, discontinuous transitions, branching-associated diversification, environmental perturbations, and imperfect experimental observations. Existing stochastic models—including Brownian motion, Ornstein–Uhlenbeck (OU) processes, Lévy jump processes, and branching processes—capture complementary aspects of these dynamics but have largely been developed as separate methodological frameworks. We present the Ornstein–Uhlenbeck–Lévy–Branching (OULB) framework, a unified probabilistic model for multiscale biological evolution that integrates continuous dynamics, discontinuous jumps, branch-associated regime dynamics, therapy-dependent attractor shifts, and observation models within a common latent-state representation. Brownian motion provides an unconstrained diffusion baseline, while classical OU, treatment-shifted OU, OU–Lévy, and OU–Branching models arise as nested or constrained formulations within the broader framework. Using a comprehensive simulation platform, we demonstrate generally reliable recovery of continuous stochastic parameters, condition-dependent recovery of jump events and branch-associated regime structure, and improved parameter estimation under sparse, noisy, and irregularly sampled observations. Independent predictive model comparison further delineates the conditions under which candidate dynamical mechanisms can be distinguished and shows reduced discrimination with sparse sampling, increased measurement noise, temporal confounding, and model misspecification. We further introduce an empirical calibration strategy that derives simulation settings from longitudinal cancer evolution data, enabling biologically grounded methodological evaluation. A modular computational workflow integrates simulation, Bayesian inference, calibration, visualization, and uncertainty quantification. The OULB framework establishes a unified mathematical and computational foundation for interpretable, extensible, and reproducible probabilistic modeling across longitudinal, spatial, and multimodal biological systems.
Authors
- Seung-Hwan Kim (ORCID: https://orcid.org/0000-0002-9397-3324)
Institutions
- Fisher College (US)
Publication Details
- Journal
- Mathematical and Computational Biology
- Published
- 2026-11-20
- DOI
- https://doi.org/10.3390/mcb1010002
- Primary Topic
- Mathematical Biology Tumor Growth
- Type
- article
- Field-Weighted Citation Impact
- 0.00