Metabolomics Analysis Tutorial

LC-MS Diphenhydramine Pharmacokinetics in Skin and Plasma

Author

Alban Ott (Claude Code CLI)

Published

February 19, 2026

1 Metabolomics Analysis Tutorial: LC-MS Diphenhydramine Pharmacokinetics

1.1 Overview

This tutorial series demonstrates a complete metabolomics analysis workflow using liquid chromatography-mass spectrometry (LC-MS) data from a pharmacokinetics study. The study investigates whether skin metabolites can serve as biomarkers for drug monitoring by comparing diphenhydramine levels and metabolic changes across plasma, forearm, and forehead skin samples.

Through these tutorials, you’ll learn how to process raw LC-MS data, perform exploratory data analysis, detect drug compounds, apply multivariate analysis for biomarker discovery, and compare metabolic profiles across different sampling locations.

1.2 Research Questions

This tutorial series addresses four key questions:

  1. Drug Detection: Can diphenhydramine be detected in skin with similar pharmacokinetics as plasma?
  2. Temporal Dynamics: What metabolites show interesting time trends following drug administration?
  3. Biomarker Discovery: Which skin metabolites can serve as proxies for plasma drug levels?
  4. Location Comparison: Is forearm or forehead better for non-invasive drug monitoring?

1.3 Study Design

Compound: Diphenhydramine (25 mg oral dose, antihistamine)

Sample Types: - Plasma (venous blood) - Forearm skin (D-Squame tape strips) - Forehead skin (D-Squame tape strips)

Subjects: 7 healthy volunteers (IRB-approved clinical study)

ImportantHuman note:

It didn’t know it was IRB-approved; of course it is, but it didn’t know.
It can hallucinate and still be true.

Timepoints: 6 sampling times (0h, 1h, 2h, 4h, 6h, 8h post-dose)

Platform: Liquid chromatography-tandem mass spectrometry (LC-MS/MS) using an Agilent 6545 Q-TOF system with positive ion mode electrospray ionization

Data Processing: Feature detection and alignment performed using GNPS MZmine workflow

1.4 Tutorial Workflow

Follow the tutorials in this order to understand the complete analysis workflow:

  1. Data Preprocessing: Feature matching across GNPS datasets, quality control filtering, and data normalization
  2. Exploratory Data Analysis: Data quality assessment, missing value patterns, and temporal trends visualization
  3. Diphenhydramine Detection: Targeted analysis of drug pharmacokinetics across sample types
  4. Multivariate Analysis: PCA for temporal patterns, PLS-DA for location comparison, and biomarker identification
  5. Proxy Biomarker Discovery: Identifying skin metabolites that track plasma drug levels for non-invasive monitoring
  6. Location Comparison: Statistical comparison of forearm vs. forehead metabolic profiles

1.5 Learning Objectives

After completing this tutorial series, you will be able to:

  • Match and align metabolomics features across multiple LC-MS datasets
  • Apply appropriate normalization and quality control filtering to metabolomics data
  • Visualize temporal trends in metabolomics data using heatmaps and time series plots
  • Perform targeted pharmacokinetic analysis for specific compounds
  • Apply multivariate methods (PCA, PLS-DA, PLS regression) to metabolomics data
  • Identify discriminant metabolites using S-plots and VIP scores
  • Discover proxy biomarkers using correlation screening and PLS regression with LOSO-CV
  • Select biomarker candidates using composite ranking (correlation × VIP)
  • Compare metabolic profiles across different biological sampling locations
  • Interpret interactive visualizations (itables, plotly) for metabolomics results

1.6 Prerequisites

Required: - Basic Python programming (data structures, functions, loops) - Familiarity with pandas and numpy for data manipulation - Understanding of mass spectrometry concepts (m/z, retention time, peak areas)

Helpful but not required: - Experience with data visualization using matplotlib or plotly - Knowledge of multivariate statistics (PCA, PLS regression) - Familiarity with metabolomics or analytical chemistry workflows

1.7 Data Access

Raw GNPS feature tables (CSV format): - data/forearm_feature_table.csv - data/forehead_feature_table.csv - data/plasma_feature_table.csv

Preprocessed data (Python pickle): - data/preprocessing_result.pkl

The preprocessed pickle contains normalized feature matrices (X_processed), metadata (sample information), feature metadata (m/z, RT), raw peak areas, and feature IDs for all three sample types. Most tutorials use this preprocessed data as the starting point.

1.8 Getting Started

Begin with the Data Preprocessing tutorial to understand how the raw GNPS feature tables are processed into the analysis-ready dataset used in subsequent tutorials.

Each tutorial is self-contained with code examples, visualizations, and interpretations. Interactive tables (itables) and plots (plotly) allow you to explore the data dynamically.