Events

R as a single-step platform for PK/PD and PBPK/QSP M&S: integration of model estimation, optimization, simulation, and reporting

October

19

2017


Thursday, October 19, 2017
8:30 AM – 5:00 PM EDT
Marriott Resort: Harbor Beach Hotel
Fort Lauderdale, FL

Summary

We’ll introduce the concept of a single platform for modeling and simulation using R. After a brief refresher on basic mrgsolve use (an R package for simulation from ODE-based PKPD and QSP models), we will lead you through case studies reflecting real-world modeling and simulation workflows. The hands-on examples will focus on planning, executing, and presenting simulation-based answers to questions in drug development.
Workshop examples will include:

  1. exploring study designs by interfacing mrgsolve with the optimal design package PopED
  2. generating MAP Bayes estimates from established population PK models
  3. demonstrating several optimizers available in R for parameter estimation in non-hierarchical (deterministic) models
In addition, we’ll review model PLUGINs to extend the mrgsolve model format for more advanced simulation models. This will all lead into an annotated model format and a process for creating modeling simulation documentation in .pdf, .html , or .md format - all on a single-source, open-source, easily-traceable and reproducible compute platform.

Materials Provided to Participants

Course slides, data and code for all examples, online access for one month to a cloud-based compute server on which the software used in the course is installed.

Requirements

Knowledge and experience in population PKPD modeling and the use of R (or S-PLUS). Previous exposure to the mrgsolve simulation package is encouraged

Course will be led by Metrum's Kyle Baron, Pharm. D. Ph.D.

Workshop Outline

  • A brief mrgsolve refresher
    • Where to find mrgsolve
    • Where to find help with mrgsolve
    • Review basic model specification elements
    • Review basic simulation workflow with mrgsolve
  • Model parameter estimation in R
    • Introduction to optimization workflow in R
      • Using stats::optim
      • Objective functions for optimization
      • The optimhelp package
    • Parameter estimation for PK/PD and QSP models
      • Organizing inputs and outputs
      • Other optimizers: minqa, RcppDE, MCMCpack
  • Generating MAP Bayes estimates in R with mrgsolve
    • Investigate the influence of residual error variance on MAP Bayes estimates
  • Interfacing mrgsolve with PopED to explore optimal designs
  • Planning & organizing simulations to answer questions in development
    • Multiple strategies for setting up simulations depending on the question and the desired output
    • Three case studies using population PK/PD and PBPK/QSP models. For each case study we will:
      • Ask questions that commonly arise in drug development
      • Develop a technical plan for simulation to address those questions
      • Simulate and present results
  • Annotated model specification and rendering model documents
  • Closing comments, questions, and answers
Register with ACOP