LLMs in the Class: A Case Study on Using Large Language Models to Teach Chemometrics with the Wisconsin Breast Cancer Dataset

This study explores the pedagogical integration of Large Language Models (LLMs) into chemistry education through a practical chemometrics activity using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. Graduate students employed Microsoft 365 Copilot (GPT-5) and Gemini to perform statistical analyses, dimensionality reduction, and classification tasks entirely via natural language prompts. The exercise covered exploratory data visualization, normality assessment, log transformation, Principal Component Analysis (PCA), and Partial Least Squares Discriminant Analysis (PLS-DA). Students compared raw and log-transformed datasets to investigate how preprocessing affected multivariate discrimination and predictive accuracy. Both LLMs generated reproducible results consistent with Jamovi software outputs and produced publication-quality plots including score, loading, VIP, and confusion matrix diagrams. Beyond technical proficiency, the activity enhanced students’ conceptual understanding of supervised and unsupervised learning while promoting critical evaluation of generative AI outputs. The findings demonstrate that LLMs can serve as accessible, interactive tools for teaching machine learning and chemometric analysis, lowering programming barriers and fostering data literacy.