Exploration of sexual dimorphism and inter-individual variability in multivariate parameter spaces for a pharmacokinetic compartment model

Megerle L. Scherholz, Ioannis Androulakis

Research output: Contribution to journalArticle

Abstract

Pharmacokinetic models are particularly useful to study the underlying and complex physiological mechanisms contributing to clinical differences across patient subgroups or special populations. Unfortunately, the inherent variability of biological systems and knowledge gaps in physiological data limit confidence in model predictions for special populations. Sourcing data to reflect the desired physiologies can be resource intensive, particularly for a larger model. Thus, a critical step in model development for special populations involves an in-depth analysis of model inputs, which can be guided by Monte Carlo simulations. Such an approach enables the generation of parameter values by stochastic sampling that are subsequently restricted to the combinations that describe biologically plausible or target model output. Our approach utilized a published pharmacokinetic compartmental model to demonstrate how sampling in conjunction with global sensitivity analysis can be used to explore sexual dimorphism and inter-individual variability in multivariate parameter spaces for differentiation of model input and behavior across phenotypes. Despite limiting the model output to relatively narrow ranges, male and female phenotypes were associated with wide variability in both individual parameter values and combinations of parameters. Through an integrated approach using a support vector machine, principal component analysis and global sensitivity analysis, our approach revealed that specific combinations of parameters gave rise to a certain phenotype, while individual parameters influenced the shape of plasma concentration profile. Augmenting analysis of the model input with global sensitivity analysis enabled an understanding of both sexual dimorphism and inter-individual variability in pharmacokinetics. While the current study revealed how model input could be separated by sex for a simple compartment model, the approach could be extended to other patient factors, such as age or disease, and to a more complex physiologically-based model that describes absorption, distribution, metabolism, and elimination with more detail.

Original languageEnglish (US)
Pages (from-to)70-80
Number of pages11
JournalMathematical Biosciences
Volume308
DOIs
StatePublished - Feb 1 2019

Fingerprint

Compartment Model
Pharmacokinetics
Sex Characteristics
sexual dimorphism
pharmacokinetics
Parameter Space
Phenotype
Population
Space Simulation
Global Analysis
Principal Component Analysis
Sensitivity Analysis
Model
Sensitivity analysis
phenotype
Compartmental Model
Confidence Limits
Output
Physiology
Sampling

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Modeling and Simulation
  • Biochemistry, Genetics and Molecular Biology(all)
  • Immunology and Microbiology(all)
  • Agricultural and Biological Sciences(all)
  • Applied Mathematics

Keywords

  • Principal component analysis
  • Sex differences
  • Special population
  • Support vector machine

Cite this

@article{4772ea0bedfb495284f9a7b09f17f6bb,
title = "Exploration of sexual dimorphism and inter-individual variability in multivariate parameter spaces for a pharmacokinetic compartment model",
abstract = "Pharmacokinetic models are particularly useful to study the underlying and complex physiological mechanisms contributing to clinical differences across patient subgroups or special populations. Unfortunately, the inherent variability of biological systems and knowledge gaps in physiological data limit confidence in model predictions for special populations. Sourcing data to reflect the desired physiologies can be resource intensive, particularly for a larger model. Thus, a critical step in model development for special populations involves an in-depth analysis of model inputs, which can be guided by Monte Carlo simulations. Such an approach enables the generation of parameter values by stochastic sampling that are subsequently restricted to the combinations that describe biologically plausible or target model output. Our approach utilized a published pharmacokinetic compartmental model to demonstrate how sampling in conjunction with global sensitivity analysis can be used to explore sexual dimorphism and inter-individual variability in multivariate parameter spaces for differentiation of model input and behavior across phenotypes. Despite limiting the model output to relatively narrow ranges, male and female phenotypes were associated with wide variability in both individual parameter values and combinations of parameters. Through an integrated approach using a support vector machine, principal component analysis and global sensitivity analysis, our approach revealed that specific combinations of parameters gave rise to a certain phenotype, while individual parameters influenced the shape of plasma concentration profile. Augmenting analysis of the model input with global sensitivity analysis enabled an understanding of both sexual dimorphism and inter-individual variability in pharmacokinetics. While the current study revealed how model input could be separated by sex for a simple compartment model, the approach could be extended to other patient factors, such as age or disease, and to a more complex physiologically-based model that describes absorption, distribution, metabolism, and elimination with more detail.",
keywords = "Principal component analysis, Sex differences, Special population, Support vector machine",
author = "Scherholz, {Megerle L.} and Ioannis Androulakis",
year = "2019",
month = "2",
day = "1",
doi = "10.1016/j.mbs.2018.12.011",
language = "English (US)",
volume = "308",
pages = "70--80",
journal = "Mathematical Biosciences",
issn = "0025-5564",
publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Exploration of sexual dimorphism and inter-individual variability in multivariate parameter spaces for a pharmacokinetic compartment model

AU - Scherholz, Megerle L.

AU - Androulakis, Ioannis

PY - 2019/2/1

Y1 - 2019/2/1

N2 - Pharmacokinetic models are particularly useful to study the underlying and complex physiological mechanisms contributing to clinical differences across patient subgroups or special populations. Unfortunately, the inherent variability of biological systems and knowledge gaps in physiological data limit confidence in model predictions for special populations. Sourcing data to reflect the desired physiologies can be resource intensive, particularly for a larger model. Thus, a critical step in model development for special populations involves an in-depth analysis of model inputs, which can be guided by Monte Carlo simulations. Such an approach enables the generation of parameter values by stochastic sampling that are subsequently restricted to the combinations that describe biologically plausible or target model output. Our approach utilized a published pharmacokinetic compartmental model to demonstrate how sampling in conjunction with global sensitivity analysis can be used to explore sexual dimorphism and inter-individual variability in multivariate parameter spaces for differentiation of model input and behavior across phenotypes. Despite limiting the model output to relatively narrow ranges, male and female phenotypes were associated with wide variability in both individual parameter values and combinations of parameters. Through an integrated approach using a support vector machine, principal component analysis and global sensitivity analysis, our approach revealed that specific combinations of parameters gave rise to a certain phenotype, while individual parameters influenced the shape of plasma concentration profile. Augmenting analysis of the model input with global sensitivity analysis enabled an understanding of both sexual dimorphism and inter-individual variability in pharmacokinetics. While the current study revealed how model input could be separated by sex for a simple compartment model, the approach could be extended to other patient factors, such as age or disease, and to a more complex physiologically-based model that describes absorption, distribution, metabolism, and elimination with more detail.

AB - Pharmacokinetic models are particularly useful to study the underlying and complex physiological mechanisms contributing to clinical differences across patient subgroups or special populations. Unfortunately, the inherent variability of biological systems and knowledge gaps in physiological data limit confidence in model predictions for special populations. Sourcing data to reflect the desired physiologies can be resource intensive, particularly for a larger model. Thus, a critical step in model development for special populations involves an in-depth analysis of model inputs, which can be guided by Monte Carlo simulations. Such an approach enables the generation of parameter values by stochastic sampling that are subsequently restricted to the combinations that describe biologically plausible or target model output. Our approach utilized a published pharmacokinetic compartmental model to demonstrate how sampling in conjunction with global sensitivity analysis can be used to explore sexual dimorphism and inter-individual variability in multivariate parameter spaces for differentiation of model input and behavior across phenotypes. Despite limiting the model output to relatively narrow ranges, male and female phenotypes were associated with wide variability in both individual parameter values and combinations of parameters. Through an integrated approach using a support vector machine, principal component analysis and global sensitivity analysis, our approach revealed that specific combinations of parameters gave rise to a certain phenotype, while individual parameters influenced the shape of plasma concentration profile. Augmenting analysis of the model input with global sensitivity analysis enabled an understanding of both sexual dimorphism and inter-individual variability in pharmacokinetics. While the current study revealed how model input could be separated by sex for a simple compartment model, the approach could be extended to other patient factors, such as age or disease, and to a more complex physiologically-based model that describes absorption, distribution, metabolism, and elimination with more detail.

KW - Principal component analysis

KW - Sex differences

KW - Special population

KW - Support vector machine

UR - http://www.scopus.com/inward/record.url?scp=85058966190&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058966190&partnerID=8YFLogxK

U2 - 10.1016/j.mbs.2018.12.011

DO - 10.1016/j.mbs.2018.12.011

M3 - Article

C2 - 30557560

AN - SCOPUS:85058966190

VL - 308

SP - 70

EP - 80

JO - Mathematical Biosciences

JF - Mathematical Biosciences

SN - 0025-5564

ER -