This repository contains a MATLAB workflow for using feature-ranking techniques as a pre-screening method for quantitative systems pharmacology (QSP) models. The main script simulates a SimBiology model across sampled parameter values, computes response summaries, ranks input parameters with multiple machine learning feature-ranking methods, and compares the rankings with saved global sensitivity analysis results.
Feature_Ranking_script.m- main MATLAB script exported from a live script. It loads the SimBiology model, generates parameter scenarios, runs simulations, computes feature-ranking scores, and plots summary results.Feature_Ranking.prj- MATLAB project file.mPBPK_siRNA.sbproj- SimBiology project containing the model used by the workflow.helper/- helper functions for selecting free model parameters, filtering parameter objects, plotting feature-ranking summaries, plotting sensitivity convergence, and setting plot defaults.results/resultsGSA.mat- saved global sensitivity analysis results used for comparison.resources/- MATLAB project metadata.
The workflow is intended for MATLAB with the following products:
- SimBiology
- Statistics and Machine Learning Toolbox
- Parallel Computing Toolbox (optional)
The script uses parallel execution for simulations and some ranking/modeling steps, but can also run as is if the Parallel Computing Toolbox is not available.
Open the MATLAB project from the repository root by double clicking on Feature_Ranking.prj or typing:
openProject("Feature_Ranking.prj")The full simulation uses Nsamples = 10^4, so runtime may vary depending on local CPU and available parallel workers.