Skip to main content
PK-Swift v2.0
In-Browser PK Textbook-compliant
Pharmacokinetic Analysis Platform

AI-Powered PK Analysis
NCA · BE · IVIVC in Your Browser

Paste any PK data — Gemini AI parses it instantly. Textbook-compliant NCA, bioequivalence assessment, IVISVC Level A, and publication-quality visualization. All PK computations run in-browser.

Gemini AI Parse Smart Parse Offline Fallback In-Browser PK Computation
NCA BE Graph NPS IVIVC Cross-Route
Validation Benchmark FAQ Disclaimer & Legal Notice
PK computations run entirely in-browser. AI Parse (optional) sends data to Google Gemini API — see data privacy notice.
Time (h) Conc (ng/mL) Cmax
5
Analysis Modules
20+
PK Parameters
AI
Gemini Parse
Zero
Install Required
BE
Bioequivalence
Analysis Modules

Six Integrated Modules

Complete pharmacokinetic analysis workflow — NCA, BE, Graph Studio, NPS, IVIVC, and Cross-Route — all client-side.

Stable v1.0

NCA Analyzer

Textbook-compliant NCA with Linear-up Log-down trapezoidal integration and automatic Best Fit λz selection.

  • AUC(0-t), AUC(0-∞), AUMC, MRT
  • CL/F, Vz/F, Vss/F, t½
  • CSV / JSON / Manual input
  • D3.js interactive visualization
New v2.0

Graph Studio

Instantly generate journal-ready formal and presentation-ready casual concentration-time graphs.

  • Formal (Journal) / Casual (Report) styles
  • Time unit conversion (min / hr / day)
  • Linear & Semi-log scale
  • SVG / PNG high-resolution export
  • Multi-profile overlay
New v1.0

NPS Superposition

Predict multiple-dose and steady-state concentrations from single-dose profiles using Nonparametric Superposition.

  • Linear-Log interpolation/extrapolation
  • Non-uniform dosing schedule
  • Steady State Cmax/Cmin/Cavg
  • Accumulation ratio & fluctuation
New v2.0

IVIVC Level A

FDA guideline-compliant Level A IVIVC analysis with Wagner-Nelson & Loo-Riegelman deconvolution.

  • Wagner-Nelson (1-compartment)
  • Loo-Riegelman (2-compartment)
  • Weibull / Hill dissolution fitting
  • Levy Plot & Fabs vs Fdiss
  • %PE Internal/External validation
New v2.0

BE Analysis

Bioequivalence assessment with per-arm NCA, geometric mean T/R ratio, 90% CI, and TOST for preclinical and clinical studies.

  • Multi-arm NCA (Test vs Reference)
  • Geomean T/R% + 90% CI (80–125%)
  • Preclinical screen & Clinical BE
  • AI Parse with auto arm detection
  • Mean ± SD & Spaghetti plots
New v2.1

Crossover BE

2×2 crossover bioequivalence with mixed-effects ANOVA, 90% CI, TOST, and within-subject CV.

  • 2×2 Crossover ANOVA (Type III SS)
  • LS Means + 90% CI (80–125%)
  • TOST & Sequence Effect Assessment
  • Within-Subject CV% & Variance Components
  • Per-Subject Data Table + Export
New v1.0

Cross-Route PK

Compare PK profiles across administration routes (oral, IV, SC, IM) with absolute bioavailability (F) and dose-normalized overlay.

  • Multi-route NCA comparison
  • Absolute bioavailability (F) calculation
  • Dose-normalized overlay chart
  • Pairwise route comparison tables
  • XLSX / PDF export
Core Features

Why PK-Swift?

Designed for regulatory science, data security, and research efficiency.

🔒
Privacy-First Design
All PK computations run in-browser. AI Parse (optional) uses Google Gemini API — review data handling policies for your API tier.
⚡
Web Workers
Parallel processing via Web Workers — heavy computations without UI blocking.
📊
Publication Quality
D3.js-based publication-ready graphs with SVG/PNG high-resolution export.
✅
Validated
Automated unit test validation against published pharmacokinetic reference values.
🧮
Precision Numerics
Floating-point error-minimized numerical analysis engine.
📐
FDA Compliance
IVIVC %PE validation criteria — mean <10%, individual <15% auto-calculated.
🔄
Deconvolution
In vivo absorption fraction via Wagner-Nelson & Loo-Riegelman.
📈
Model Fitting
Weibull/Hill nonlinear regression dissolution profile fitting engine.
🤖

AI Parse — Paste Anything

Powered by the latest Google Gemini model, PK-Swift's AI Parse understands any format you throw at it — messy Excel pastes, PDFs, clinical reports, lab notebooks, or even free-text descriptions. It auto-detects groups, subjects, arms, and time-concentration columns. No reformatting needed.

Excel / CSV / TSV paste
JSON & structured data
Multi-subject / Multi-arm
BLQ/BQL → auto zero
Natural language input
Offline Smart Parse fallback
Validation

Live Benchmark — Theophylline Dataset

Auto-validated NCA computation against published textbook reference values. Score updates on every page load.

— Accuracy

Benchmark Summary

Tests—
Passed—
Failed—
Geo. Mean Acc.—
DatasetTheophylline Subj.1
MethodLin-up Log-down

Standard Dataset

Theophylline Subject 1
Dose: 319.992 mg (oral, extravascular)
11 time points (0 – 24.37 h)
Linear-up Log-down trapezoidal rule
Best-fit λz terminal regression
📖 Gabrielsson & Weiner, Pharmacokinetic & Pharmacodynamic Data Analysis, 5th ed. §10.2
📖 Rowland & Tozer, Clinical Pharmacokinetics, 4th ed. Ch.4
📖 Gibaldi & Perrier, Pharmacokinetics, 2nd ed. §6
Parameter Expected Computed |Δ| Tolerance Accuracy Result Reference
Computing benchmark...
Tolerance Criteria & Regulatory References

Each test parameter is validated against published textbook reference values. PASS criteria are based on absolute tolerance (|Computed − Expected| ≤ Tolerance), calibrated to account for method-dependent variation commonly observed across validated NCA software platforms.

Observed values (Cmax, Tmax, Clast, Tlast): Strict — near-zero tolerance. Direct read from data.
AUC parameters: ≤0.5–2.0 units. Integration method variation (linear vs. log-linear) accounts for typical ±1% differences.
Terminal phase (λz, t½): ≤0.5–2.0 units. λz regression point selection (Best-fit vs. manual) commonly introduces ±1–5% variation across platforms (Gabrielsson §10.3.2).
Derived parameters (CL/F, Vz/F, Vss/F, MRT): Dose-dependent; tolerance scaled to expected magnitude. Variation propagated from AUCinf and λz estimates.
Quality metrics (R²adj, AUC%extrap): Dimensionless. R²adj ≥0.99 required for reliable λz; AUC%extrap <20% per FDA BE guidance.
Cross-validation: All-linear AUC computed independently to verify log-linear method consistency.

Key References: [1] Gabrielsson & Weiner, PK/PD Data Analysis, 5th ed., 2016 — [2] Rowland & Tozer, Clinical Pharmacokinetics, 4th ed., 2011 — [3] Gibaldi & Perrier, Pharmacokinetics, 2nd ed., 1982 — [4] Wagner, Fundamentals of Clinical Pharmacokinetics, 1975 — [5] Zhang et al., PKSolver, Comput Methods Programs Biomed, 2010

Open Full Benchmark Page →