动态贝叶斯网络驱动的抗体药物警戒范式探索：从静态因果到时序概率网络-现代医学

动态贝叶斯网络驱动的抗体药物警戒范式探索：从静态因果到时序概率网络

单位：国家药品监督管理局药品审评检查长三角分中心, 上海 200120

分类号：R186

出版年·卷·期（页码）：2025·53·第十一期（1703-1710）

摘要：

目的：探索基于动态贝叶斯网络(DBN)的抗体药物警戒监测新范式。方法：整合暴露剂量、遗传多态性、生物标志物及临床表型等多维动态数据，构建包含9个关键节点的DBN时序概率网络，建立不良反应风险的多时间片传导模型。采用蒙特卡洛近似推断与时间序列动态推断实现风险预测。结果：构建的DBN模型实现了不良反应风险的动态量化与个体化预测。仿真研究显示，模型能够根据患者个体特征动态预测不良反应概率，并追踪信号强度的时序演变规律。结论：基于DBN的动态风险预测框架可为抗体药物不良反应防控提供个性化决策支持，为新型生物制品的风险动态管控提供技术路径借鉴。

Objective: To explore a new paradigm for antibody pharmacovigilance monitoring using dynamic Bayesian network(DBN). Methods: By integrating multidimensional data including dose exposure, genetic polymorphisms, biomarkers, and clinical phenotypes, a DBN-based temporal probability network comprising nine key nodes was constructed to establish a multi-time-slice propagation model for adverse drug reaction(ADR) risks. Monte Carlo approximate inference and temporal dynamic inference were employed to achieve risk prediction. Results: The constructed DBN model achieved dynamic quantification and individualized prediction of ADR risks. Simulation studies demonstrated that the model could dynamically predict ADR probabilities based on individual patient characteristics and track the temporal evolution patterns of signal intensity. Conclusion: The DBN-based dynamic risk prediction framework can provide personalized decision support for preventing antibody-drug ADRs and offer a practical pathway for managing the dynamic risk of novel biological products.

参考文献：

[1] ALMENOFF J,TONNING J M,GOULD A L,et al.Perspectives on the use of data mining in pharmaco-vigilance[J].Drug Saf,2005,28(11):981-1007.
[2] BATE A,STEGMANN J U.Safety of medicines and vaccines-building next generation capability[J].Trends Pharmacol Sci,2021,42(12):1051-1063.
[3] RUDIN C.Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead[J].Nat Mach Intell,2019,1(5):206-215.
[4] HAUSER P,TAN X,CHEN F,et al.Improved pharmacovigilance signal detection using Bayesian generalized linear mixed models[J].Stat Med,2025,44(10-12):e70086.
[5] MURPHY K.Dynamic Bayesian networks:representation,inference and learning[M].2002.
[6] BENDER S,GAMERMAN V,REESE P P,et al.The first-order Markov conditional linear expectation approach for analysis of longitudinal data[J].Stat Med,2021,40(8):1972-1988.
[7] GHANMI N,ALI M,ESSOUKRI N.Characterization of dynamic Bayesian network-the dynamic Bayesian network as temporal network[J].Int J Adv Comput Sci Appl,2011,2(7):53-60.
[8] ALQUIER P.Approximate Bayesian inference[J].Entropy,2020,22(11):1272.
[9] SUCAR L E.Dynamic and temporal Bayesian networks[M]//Probabilistic Graphical Models:Principles and Applications.Cham:Springer International Publishing,2021:181-202.
[10] MARTIN O.Bayesian analysis with python:introduction to statistical modeling and probabilistic programming using PyMC3 and ArviZ[M].Birmingham:Packt Publishing Ltd,2018:143-177.
[11] HANSEL T T,KROPSHOFER H,SINGER T,et al.The safety and side effects of monoclonal antibodies[J].Nat Rev Drug Discov,2010,9(4):325-338.
[12] SHANKAR G,ARKIN S,COCEA L,et al.Assessment and reporting of the clinical immunogenicity of therapeutic proteins and peptides-harmonized terminology and tactical recommendations[J].AAPS J,2014,16(4):658-673.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】

提示：您还未登录，请登录！点此登录