Frequently Asked Questions

PK-Swift is a browser-based pharmacokinetic (PK) analysis platform. It provides Non-Compartmental Analysis (NCA), Bioequivalence (BE) assessment, NPS Superposition, IVIVC Level A analysis, and publication-quality graph generation — all running directly in your web browser without any installation or server backend.

No. PK-Swift runs entirely in your web browser. Just open the page and start analyzing. No downloads, no plugins, no server setup required.

Yes. PK-Swift itself is free. The core PK computation features (NCA, BE, NPS, IVIVC, Graph Studio) require no API key or subscription. The optional AI Parse feature uses the Google Gemini API, which requires a free API key from Google AI Studio.

PK-Swift works best on modern browsers — Chrome (recommended), Firefox, Safari, and Edge. It requires JavaScript enabled and supports Web Workers for parallel computation. Mobile browsers are supported with responsive layouts.

PK computations (NCA, NPS, IVIVC, BE, and Graph Studio) run entirely within your browser using JavaScript. Your pharmacokinetic data is not transmitted to any external server for these features.

However, the optional AI Parse feature sends the text you paste to Google's Gemini API for intelligent parsing. If data privacy is a critical concern, use the Smart Parse (offline rule-based parser) instead, which sends nothing externally.

When you use AI Parse:

• The text you paste is sent to Google's Gemini API servers.
• Free-tier API users: Google may use free-tier input/output to improve their models. Check Google's Gemini API Terms.
• Paid-tier API users: Typically have different (more restrictive) data usage terms — consult your Google Cloud agreement.
• Recommendation: Do NOT paste patient-identifiable, confidential, or proprietary data into AI Parse unless you have reviewed Google's policies for your API tier.

Your API key is stored only in your browser's localStorage. PK-Swift never transmits your API key to any server other than Google's Gemini API endpoint during AI Parse requests. You can clear it at any time in Settings (⚙).

• Smart Parse — A built-in, offline, rule-based parser. It handles CSV, TSV, JSON, Excel paste, semicolon-delimited, and some natural language patterns. No API key needed. No data sent externally. Always available.
• AI Parse — Uses Google's Gemini API for complex or messy data formats that rule-based parsing cannot handle. Requires a Gemini API key. Data is sent to Google for processing.

PK-Swift uses the Linear-up Log-down trapezoidal rule, which is the most widely used method in commercial NCA software. Ascending concentration segments use the linear trapezoidal formula, and descending segments use the log-trapezoidal formula. This is described in Gabrielsson & Weiner (2016) and Gibaldi & Perrier (1982).

PK-Swift uses a "Best Fit" algorithm: it considers all subsets of ≥ 3 terminal data points (post-Tmax, concentration > 0), requires a negative slope, and selects the subset with the highest adjusted R². Ties are broken by choosing the largest number of points. This approach is consistent with the methodology used in Phoenix WinNonlin and other NCA software.

PK-Swift includes an automated Validation Benchmark that computes 18 NCA parameters from the Theophylline Subject 1 standard dataset and compares them against published reference values. The benchmark runs automatically on the landing page. Typical geometric mean accuracy exceeds 99%. Users can also run the full benchmark to verify computations independently.

No. PK-Swift is intended for educational, research, and exploratory purposes only. It is NOT a validated software system under 21 CFR Part 11 or any regulatory framework. For regulatory submissions, use commercially validated PK software with established SOPs and audit trails. See the Disclaimer for full details.

The BE (Bioequivalence) module allows comparison of Test vs. Reference arms. It computes per-arm NCA parameters, geometric mean T/R ratios, and 90% confidence intervals for Cmax and AUClast. For preclinical studies, it provides screening against the 80–125% bioequivalence criteria. For clinical studies, it reports whether the 90% CI falls within [80%, 125%] for a BE conclusion. Note: full ANOVA-based BE analysis (e.g., 2×2 crossover) is not yet implemented.

PK-Swift accepts a wide variety of input formats:

• CSV (comma-separated), TSV (tab-separated)
• Excel copy-paste (with or without headers)
• JSON arrays of {time, conc} objects
• Semicolon-delimited data
• Natural language (e.g., "At 0 hours the concentration was 0.74 ng/mL...")
• Multi-subject / multi-arm wide-format tables

Smart Parse handles most standard formats offline. For messy or unusual formats, use AI Parse.

• NCA Analyzer: Export results as CSV, JSON, or formatted text report.
• Graph Studio: Export charts as SVG (vector) or PNG (high-resolution).
• NPS / IVIVC: Export computed data as XLSX spreadsheets.

Yes. In Graph Studio, use the style toggle buttons at the top. Formal produces journal-ready black-and-white graphs with serif fonts, ideal for publications. Casual produces colorful, modern graphs with gradient fills, suited for presentations and reports. The Spaghetti plot also adapts to the selected style.

Visit Google AI Studio and sign in with your Google account. Generate a free API key. Then open PK-Swift's Settings (⚙ icon) on any page and paste your key. It's saved in your browser only.

Yes, PK-Swift has responsive layouts and works on tablets and smartphones. However, for the best experience with data tables and graph interaction, a desktop or laptop screen is recommended. The Graph Studio's SVG export and spaghetti plots are optimized for larger screens.

Try these steps:

• Ensure your data has clear time and concentration columns (headers like "Time", "Conc", "hr", "ng/mL" help Smart Parse).
• If Smart Parse fails, try AI Parse — it handles messy formats much better.
• Check for non-numeric values in concentration fields. BLQ/BQL entries are automatically converted to 0.
• If your data is in a wide format (multiple subjects as columns), AI Parse can auto-detect this structure.
• As a last resort, you can enter data manually point by point.

Absolutely — this is a recommended use case. PK-Swift implements the same standard Linear-up Log-down method used by major NCA software. Small numerical differences (typically <1%) may occur due to λz point selection algorithms and floating-point precision. The built-in Validation Benchmark demonstrates agreement with published reference values. Cross-validation against validated software is encouraged for research purposes.

PK-Swift's IVIVC Level A module implements Wagner-Nelson and Loo-Riegelman deconvolution with %PE validation criteria (mean <10%, individual <15%) consistent with FDA guidance. However, it is NOT a validated regulatory tool. For regulatory submissions, use commercially validated software with full audit trails and SOPs. PK-Swift is suitable for exploratory IVIVC analysis, educational purposes, and preliminary feasibility assessment.