Stage-specific evolutionary dynamics in pediatric leukemia inferred using Bayesian stochastic modeling
- 1Department of Biology, Fisher College, Boston, Massachusetts, United States of America
Abstract
Tumor evolution in pediatric leukemia is shaped by stochastic fluctuations and constraints imposed by developmental programs and therapeutic pressure. However, quantitative characterization of these dynamics remains challenging because of sparse longitudinal sampling and measurement noise in clinical data. In this study, we applied Bayesian calibration of stochastic Ornstein–Uhlenbeck (OU) models to characterize patient-specific tumor evolution trajectories in pediatric leukemia. Using a longitudinal cohort of KMT2A-rearranged leukemias (52 patients, 311 longitudinal observations corresponding to 294 unique patient-time points), we estimated parameters governing mean reversion and stochastic variability and assessed the stability and uncertainty of these estimates through simulation-based validation under empirically derived sampling conditions. A representative single-patient trajectory further illustrated patient-level evolutionary dynamics. Across the cohort, inferred OU parameters showed stage-associated tendencies, including higher average diffusion in refractory and very-early disease groups, although posterior intervals overlapped substantially and did not support sharply separated dynamical regimes. Simulation analyses indicated that Bayesian calibration provided stable posterior computation and interpretable uncertainty quantification under sparse sampling, although recovery remained parameter- and schedule-dependent and predictive performance was limited. These findings show that Bayesian-calibrated stochastic models provide a useful framework for interpreting longitudinal tumor evolution in pediatric leukemia. While not intended for clinical prediction, this approach offers a quantitative perspective on stage-dependent disease dynamics and highlights the importance of uncertainty-aware modeling in sparse clinical datasets.
Related
- Fusion-specific co-mutational architecture and treatment-response associations in pediatric core-binding factor acute myeloid leukemia: a multicenter exome-profiled cohort
- 179 Machine learning-based prediction of CAR-T cell therapy response in pediatric acute lymphoblastic leukemia patients using single-cell multi-omics features
- Sex- and age-stratified pharmacovigilance of methotrexate-associated adverse events in pediatric leukemia: a FAERS-based study
- Case Report: Dasatinib-associated chylothorax and occult Bartonella henselae infection in pediatric leukemia: a case report and systematic review
- Child Opportunity Index and Survival After Pediatric Hematopoietic Cell Transplantation for Acute Leukemia
BibTeX
@article{kim2026stagespecific,
title = {{Stage-specific evolutionary dynamics in pediatric leukemia inferred using Bayesian stochastic modeling}},
author = {Seung-Hwan Kim},
journal = {PLOS One},
year = {2026},
month = sep,
doi = {10.1371/journal.pone.0357755},
url = {https://doi.org/10.1371/journal.pone.0357755},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Stage-specific evolutionary dynamics in pediatric leukemia inferred using Bayesian stochastic modeling
AU - Seung-Hwan Kim
PY - 2026
DA - 2026/09/11
JO - PLOS One
DO - 10.1371/journal.pone.0357755
UR - https://doi.org/10.1371/journal.pone.0357755
AB - Tumor evolution in pediatric leukemia is shaped by stochastic fluctuations and constraints imposed by developmental programs and therapeutic pressure. However, quantitative characterization of these dynamics remains challenging because of sparse longitudinal sampling and measurement noise in clinical data. In this study, we applied Bayesian calibration of stochastic Ornstein–Uhlenbeck (OU) models to characterize patient-specific tumor evolution trajectories in pediatric leukemia. Using a longitudinal cohort of KMT2A-rearranged leukemias (52 patients, 311 longitudinal observations corresponding to 294 unique patient-time points), we estimated parameters governing mean reversion and stochastic variability and assessed the stability and uncertainty of these estimates through simulation-based validation under empirically derived sampling conditions. A representative single-patient trajectory further illustrated patient-level evolutionary dynamics. Across the cohort, inferred OU parameters showed stage-associated tendencies, including higher average diffusion in refractory and very-early disease groups, although posterior intervals overlapped substantially and did not support sharply separated dynamical regimes. Simulation analyses indicated that Bayesian calibration provided stable posterior computation and interpretable uncertainty quantification under sparse sampling, although recovery remained parameter- and schedule-dependent and predictive performance was limited. These findings show that Bayesian-calibrated stochastic models provide a useful framework for interpreting longitudinal tumor evolution in pediatric leukemia. While not intended for clinical prediction, this approach offers a quantitative perspective on stage-dependent disease dynamics and highlights the importance of uncertainty-aware modeling in sparse clinical datasets.
SN - 1932-6203
ER -