Speaker
Description
Drug-induced damage to the blood-forming system, also called
hematotoxicity, is a frequent side effect of cytotoxic chemotherapy. Due
to high patient heterogeneity, it remains difficult to predict
individual treatment responses. Mechanistic models describing
thrombopoiesis provide some physiological interpretability but often
fail to capture individual irregular patient trajectories. Here, we
investigate hybrid mechanistic and data-driven approaches for
individualized prediction of platelet dynamics during chemotherapy. For
this purpose, we consider hybrid models that combine mechanistic
myelosuppression models with neural networks in a universal differential
equation framework. In addition, we present a purely data-driven
alternative, based on nonlinear auto-regressive exogenous models with
gated recurrent units. We systematically compare the approaches with
several mechanistic models across a range of real patient scenarios with
varying levels of toxicity risk, data availability and data sparsity.
Our results show that data-driven models substantially improve
predictive accuracy if sufficient longitudinal data are available,
particularly for high-risk patients with irregular platelet dynamics. In
contrast, mechanistic and hybrid approaches outperform purely
data-driven models in sparse-data regimes. These findings provide
practical guidance on modeling choices for different individual
scenarios to support clinical decision-making in chemotherapy management.
Bibliography
@article{steinacker_predicting_2024,
title = {Predicting chemotherapy-induced thrombotoxicity by {NARX} neural networks and transfer learning},
volume = {150},
issn = {1432-1335},
url = {https://link.springer.com/10.1007/s00432-024-05985-y},
doi = {10.1007/s00432-024-05985-y},
abstract = {Abstract
Background
Thrombocytopenia is a common side effect of cytotoxic chemotherapies, which is often dose-limiting. Predicting an individual’s risk is of high clinical importance, as otherwise, a small subgroup of patients limits dosages for the overall population for safety reasons.
Methods
We aim to predict individual platelet dynamics using non-linear auto-regressive networks with exogenous inputs (NARX). We consider different architectures of the NARX networks, namely feed-forward networks (FNN) and gated recurrent units (GRU). To cope with the relative sparsity of individual patient data, we employ transfer learning (TL) approaches based on a semi-mechanistic model of hematotoxicity. We use a large data set of patients with high-grade non-Hodgkin’s lymphoma to learn the respective models on an individual scale and to compare prediction performances with that of the semi-mechanistic model.
Results
Of the examined network models, the NARX with GRU architecture performs best. In comparison to the semi-mechanistic model, the network model can result in a substantial improvement of prediction accuracy for patients with irregular dynamics, given well-spaced measurements. TL improves individual prediction performances.
Conclusion
NARX networks can be utilized to predict an individual’s thrombotoxic response to cytotoxic chemotherapy treatment. For reasonable model learning, we recommend at least three well-spaced measurements per cycle: at baseline, during the nadir phase and during the recovery phase. We aim at generalizing our approach to other treatment scenarios and blood lineages in the future.},
language = {en},
number = {10},
urldate = {2025-02-19},
journal = {Journal of Cancer Research and Clinical Oncology},
author = {Steinacker, Marie and Kheifetz, Yuri and Scholz, Markus},
month = oct,
year = {2024},
pages = {457},
}
@misc{steinacker_developing_2025,
title = {Developing hybrid mechanistic and data-driven personalized prediction models for platelet dynamics},
copyright = {Creative Commons Attribution 4.0 International},
url = {https://arxiv.org/abs/2505.21204},
doi = {10.48550/ARXIV.2505.21204},
abstract = {Hematotoxicity, drug-induced damage to the blood-forming system, is a frequent side effect of cytotoxic chemotherapy and poses a significant challenge in clinical practice due to its high inter-patient variability and limited predictability. Current mechanistic models often struggle to accurately forecast outcomes for patients with irregular or atypical trajectories. In this study, we develop and compare hybrid mechanistic and data-driven approaches for individualized time series modeling of platelet counts during chemotherapy. We consider hybrid models that combine mechanistic models with neural networks, known as universal differential equations. As a purely data-driven alternative, we utilize a nonlinear autoregressive exogenous model using gated recurrent units as the underlying architecture. These models are evaluated across a range of real patient scenarios, varying in data availability and sparsity, to assess predictive performance. Our findings demonstrate that data-driven methods, when provided with sufficient data, significantly improve prediction accuracy, particularly for high-risk patients with irregular platelet dynamics. This highlights the potential of data-driven approaches in enhancing clinical decision-making. In contrast, hybrid and mechanistic models are superior in scenarios with limited or sparse data. The proposed modeling and comparison framework is generalizable and could be extended to predict other treatment-related toxicities, offering broad applicability in personalized medicine.},
urldate = {2025-06-02},
publisher = {arXiv},
author = {Steinacker, Marie and Kheifetz, Yuri and Scholz, Markus},
year = {2025},
note = {Version Number: 1},
keywords = {FOS: Computer and information sciences, Machine Learning (cs.LG), FOS: Biological sciences, I.6.5; J.3, Quantitative Methods (q-bio.QM)},
}
