Statistical inference of the cellular origin of chronic myeloid leukemia using a discrete-parameter ABC–PMC framework

Chronic myeloid leukemia (CML) arises from the BCR::ABL1 fusion gene, but the exact stage of cellular differentiation at which the first leukemic cell emerges remains uncertain. We develop a stochastic 27-compartment model of hematopoiesis (blood cell development) using a continuous-time multitype branching process to capture the dynamics of both healthy and cancer cells. To infer the origin of CML, we develop a discrete-parameter Approximate Bayesian Computation – Population Monte Carlo (ABC–PMC) algorithm, tailored to estimate the posterior distribution for the stage of differentiation at which the first cancer cell appeared. Applied to patient data, our method consistently identifies the stem cell compartment as the most likely source of CML. These findings improve understanding of disease initiation and demonstrate the power of discrete-parameter ABC–PMC for statistical inference in complex biological systems.