In-Silico Clinical Trials with NLME Models

A Julia/Pumas implementation of the In Silico Clinical Trial (ISCT) workflow from:

“A Step-by-Step Workflow for Performing In Silico Clinical Trials With Nonlinear Mixed Effects Models” Cortés-Ríos et al., CPT: Pharmacometrics & Systems Pharmacology (2025) DOI: 10.1002/psp4.70122

Workflow Overview

The tutorials on this site walk through the six steps of the ISCT workflow:

1. Model introduction — Define the pharmacometric or QSP model in Pumas
2. Global sensitivity analysis — Identify influential parameters via Sobol and eFAST methods
3. Structural identifiability analysis — Assess whether model parameters are identifiable from outputs
4. Copula-based virtual population generation — Sample plausible parameter sets using Gaussian copulas
5. MILP calibration — Calibrate the virtual population to clinical distributions using mixed-integer linear programming
6. Virtual clinical trial simulation — Run multi-arm VCT simulations with the calibrated population

Case Studies

Tumor Burden Model

A 3-parameter tumor burden model used as an introductory example:

• Model Introduction
• Global Sensitivity Analysis
• Structural Identifiability Analysis
• Copula Virtual Population
• MILP-Based Virtual Population Calibration
• Virtual Clinical Trial Simulation

HBV QSP Model

An 11-ODE hepatitis B virus QSP model that demonstrates the workflow on a more complex system:

• Model Introduction
• Global Sensitivity Analysis
• Structural Identifiability Analysis
• Copula Virtual Population
• MILP-Based Virtual Population Calibration
• Virtual Clinical Trial Simulation