KANACKI-PK: A Methodologically Transparent Evaluation of Symbolic Regression for Population Pharmacokinetic Modeling

This repository contains the complete synthetic and reference datasets, and raw numerical results for the study evaluating KANACKI, a deterministic symbolic regression engine, on a population pharmacokinetic (PopPK) task using the Theophylline dataset. The repository includes: Experiment 1–5 implementation scripts (fixed-effects baseline, unconstrained per-subject symbolic fit, structured one-compartment per-subject fit, shrinkage approximation, leave-one-patient-out cross-validation) The Theophylline dataset (per-subject and pooled formats), sourced from Boeckmann, Sheiner & Beal (1994) and the R nlme package Raw per-experiment output logs and result tables corresponding to Tables 1–3 in the accompanying manuscript Figure-generation scripts for the goodness-of-fit (Figure 1) and leave-one-patient-out cross-validation (Figure 2) plots This repository does not include the core KANACKI symbolic regression engine itself, which is released as an independent tool at DOI 10.5281/zenodo.20460024.