Pharmacometrics with Sounio

Sounio provides a powerful platform for pharmacometric modeling, enabling drug dosage optimization and clinical trial design with rigorous uncertainty handling.

Why Sounio for Pharmacometrics?

Traditional pharmacometric tools often treat model parameters as fixed values. In reality:

Drug absorption rates vary between patients
Measurement equipment has finite precision
Population studies have sampling uncertainty

Sounio’s epistemic types naturally capture this uncertainty.

Example: PK/PD Modeling

use std::units::{mg, L, h, mg_per_L}
use std::epistemic::{Knowledge, measure}

struct PKParameters {
    clearance: Knowledge<L/h>,
    volume: Knowledge<L>,
    ka: Knowledge<1/h>,  // Absorption rate
}

fn predict_concentration(
    dose: Knowledge<mg>,
    params: &PKParameters,
    time: h
) -> Knowledge<mg_per_L> {
    let cl = params.clearance
    let v = params.volume
    let ka = params.ka

    // One-compartment model with first-order absorption
    let ke = cl / v
    let concentration = (dose * ka) / (v * (ka - ke)) *
        (exp(-ke * time) - exp(-ka * time))

    concentration  // Uncertainty propagates automatically
}

Confidence-Gated Dosing

fn recommend_dose(
    patient: &Patient,
    target: Knowledge<mg_per_L>
) -> DoseRecommendation with IO {
    let predicted = simulate_dosing(patient, proposed_dose)

    if predicted.confidence > 0.90 {
        if within_therapeutic_window(predicted, target) {
            DoseRecommendation::Approve(proposed_dose)
        } else {
            DoseRecommendation::Adjust(calculate_adjustment(predicted, target))
        }
    } else {
        // Uncertainty too high - request more data
        perform IO::warn("Insufficient confidence for dosing recommendation")
        DoseRecommendation::RequestData(needed_measurements(patient))
    }
}

Population Modeling

fn population_pk_analysis(
    patients: Vec<PatientData>
) -> PopulationParameters with IO, Prob {
    // Hierarchical Bayesian model
    let population_cl = perform Prob::sample(LogNormal(log(10.0), 0.3))
    let population_v = perform Prob::sample(LogNormal(log(50.0), 0.2))

    for patient in patients {
        // Individual parameters with random effects
        let individual_cl = perform Prob::sample(
            LogNormal(log(population_cl), omega_cl)
        )

        // Condition on observed concentrations
        for obs in patient.observations {
            let predicted = predict_concentration(
                patient.dose, individual_cl, patient.v, obs.time
            )
            perform Prob::observe(
                Normal(predicted.value, obs.uncertainty),
                obs.concentration
            )
        }
    }

    PopulationParameters {
        clearance: population_cl,
        volume: population_v,
    }
}

Features for Pharma

GUM-compliant uncertainty: Follows ISO/GUM standards
Unit safety: Prevents mg/kg vs mg/L errors
Provenance tracking: Full audit trail for FDA compliance
Effect isolation: Separates pure computation from I/O

Regulatory Compliance

Sounio’s provenance tracking supports:

FDA 21 CFR Part 11
GxP validation requirements
ICH E9 statistical principles
Full audit trails

Get Started

sounio new pharma-project --template pharmacometrics
cd pharma-project
sounio run examples/pk_model.sio

Pharmacometrics