Enabling Green Analytical Chemistry: An Automated Chemometric Framework for Spectrophotometric Quantification of Veterinary Drugs

Effective pain management in veterinary medicine often utilizes multi-drug formulations containing Metamizole sodium (MT), Paracetamol (PA), and Diclofenac (DI). However, severe spectral overlap makes their simultaneous quantification by spectrophotometry challenging, often requiring less green chromatographic methods. This study introduces AutoRegress, a novel Automated Machine Learning (AutoML) framework designed to overcome this limitation by automating the entire chemometric model development workflow. AutoRegress orchestrates a competitive evaluation of diverse regression algorithms and feature selection techniques, revealing that the optimal analytical model is analyte-specific-a key insight challenging the conventional 'one-size-fits-all' approach. The framework objectively identified a Lasso regression for MT, a non-linear Support Vector Regression (SVR) for PA, and a distinct Lasso model for DI. These tailored models demonstrated exceptional predictive accuracy (R² > 0.988) on an independent test set. By providing the robust computational solution needed to deconvolve the complex spectra, AutoRegress enables the use of a rapid, green analytical method. The primary contribution is a reproducible platform that automates complex model selection, provides deeper chemometric insights, and makes advanced analytical solutions more accessible for routine quality control.