Learning pharmacokinetic models for in vivo glucocorticoid activation

JOURNAL OF THEORETICAL BIOLOGY

卷:455

Article

Bunte, Kerstin^1,2  Smith, David J.^3,7  Chappell, Michael J.⁴  Hassan-Smith, Zaki K.^5,6,8,10  Tomlinson, Jeremy W.⁹  Arlt, Wiebke^7,8  Tino, Peter^1,7 

[1] Univ Birmingham, Sch Comp Sci, Birmingham B15 2TT, W Midlands, England

[2] Univ Groningen, Fac Sci & Engn, POB 407, NL-9700 AK Groningen, Netherlands

[3] Univ Birmingham, Sch Math, Birmingham B15 2TT, W Midlands, England

[4] Univ Warwick, Sch Engn, Coventry CV4 7AL, W Midlands, England

[5] Queen Elizabeth Hosp Birmingham, Dept Endocrinol, Birmingham B15 2TH, W Midlands, England

[6] Queen Elizabeth Hosp Birmingham, Dept Acute Internal Med, Birmingham B15 2TH, W Midlands, England

[7] Univ Birmingham, Inst Metab & Syst Res, Birmingham, W Midlands, England

[8] Birmingham Hlth Partners, Ctr Endocrinol Diabet & Metab, Queen Elizabeth Hosp Birmingham, Birmingham, W Midlands, England

[9] Univ Oxford, Oxford Ctr Diabet Endocrinol & Metab, NIHR Oxford Biomed Res Ctr, Oxford, England

[10] Coventry Univ, Ctr Appl Biol & Exercise Sci, Coventry, W Midlands, England

关键词: Dynamic systems;    Pharmacokinetics;    Identifiability analysis;    Perturbation analysis;    11 beta-HSD activity;    In vivo glucocorticoid activation;    Probabilistic models;    Gaussian mixture model;    Expectation maximization;    Clustering;    Partially observed time series analysis;   

DOI  :  10.1016/j.jtbi.2018.07.025

来源: Elsevier

【 摘 要 】

To understand trends in individual responses to medication, one can take a purely data-driven machine learning approach, or alternatively apply pharmacokinetics combined with mixed-effects statistical modelling. To take advantage of the predictive power of machine learning and the explanatory power of pharmacokinetics, we propose a latent variable mixture model for learning clusters of pharmacokinetic models demonstrated on a clinical data set investigating 11 beta-hydroxysteroid dehydrogenase enzymes (11 beta -HSD) activity in healthy adults. The proposed strategy automatically constructs different population models that are not based on prior knowledge or experimental design, but result naturally as mixture component models of the global latent variable mixture model. We study the parameter of the underlying multi-compartment ordinary differential equation model via identifiability analysis on the observable measurements, which reveals the model is structurally locally identifiable. Further approximation with a perturbation technique enables efficient training of the proposed probabilistic latent variable mixture clustering technique using Estimation Maximization. The training on the clinical data results in 4 clusters reflecting the prednisone conversion rate over a period of 4 h based on venous blood samples taken at 20-min intervals. The learned clusters differ in prednisone absorption as well as prednisone/prednisolone conversion. In the discussion section we include a detailed investigation of the relationship of the pharmacokinetic parameters of the trained cluster models for possible or plausible physiological explanation and correlations analysis using additional phenotypic participant measurements. (C) 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license.

【 授权许可】

Free