Pharmacokinetics & Drug Dosing
Dosing a drug correctly means knowing a patient's clearance — and in real life, nobody ever measures that directly. What a clinician sees is a handful of blood draws with assay noise and a lower limit below which the lab simply reports "undetectable." Every dataset in this category is built from a real pharmacokinetic compartmental model, which means the true clearance, the true concentration curve, and the true correct dose are all known exactly, alongside the noisy, censored version a clinician would actually see.
The organ-impairment dataset here couples a two-compartment PK model to real kidney- and liver-function dynamics — augmented renal clearance, acute kidney injury onset and recovery — and checks the result against published clinical pharmacokinetics: renal clearance correlates with creatinine clearance at R²=0.76, squarely inside the 0.51–0.79 band reported for gentamicin across nearly a thousand patients.
What you'll typically find here
- Ground-truth clearance datasets — the true pharmacokinetic parameters for every patient, something no real clinical dataset can offer, alongside a realistic sparse, noisy, censored observed layer.
- Optimal-dosing labels — the correct dose for each patient's actual physiology, as a direct target variable for dosing algorithms.
- Below-limit-of-quantification data — assay censoring modelled explicitly, for testing how a method handles data it can't fully see.
Where the model diverges from the literature — and it does, in specific, named ways — the passport says so, with the size of the gap stated as a number, not a shrug.